|Exam Name||:||IBM Algo Financial Modeler Developer Fundamentals|
|Questions and Answers||:||60 Q & A|
|Updated On||:||March 20, 2018|
|PDF Download Mirror||:||C2020-002 Dump|
|Get Full Version||:||Pass4sure C2020-002 Full Version|
IBM<b><sup>®</sup></b></span> Algo Financial <span id="spanHghlt4a51">Modeler<b><sup>®</sup></b></span>, an actuarial, risk and financial modeling software solution suite.</span>" data-reactid="12">REIGATE, united kingdom , July three, 2017 /PRNewswire/ -- RNA Analytics today announced it has got the property and know-how of IBM® Algo economic Modeler®, an actuarial, chance and monetary modeling software answer suite.
IBM<b><sup>®</sup></b></span> Algo Financial <span id="spanHghlt05f3">Modeler<b><sup>®</sup></b></span> clients will continue to be supported in RNA Analytics by the transitioning team, many of whom have been working with Algo Financial <span id="spanHghlte321">Modeler<b><sup>®</sup></b></span> since its original release in 2006. Existing license users can continue to use the solution as before and will receive further announcements as the solution suite is transitioned into RNA Analytics' support structure.</span>" data-reactid="13">current IBM® Algo economic Modeler® clients will continue to be supported in RNA Analytics via the transitioning team, a lot of whom have been working with Algo financial Modeler® given that its normal unlock in 2006. latest license clients can continue to use the solution as earlier than and may receive extra announcements as the solution suite is transitioned into RNA Analytics' guide constitution.
"As at all times, our key focus is on the shoppers and their requirements, this chance will permit us to be aware of the service and utility innovation required to deliver purchasers with a market leading analytics providing for the actuarial and possibility management services. The shoppers the use of Algo fiscal Modeler® will proceed to improvement from the superior solution facets and our expanding consulting functions all over the world" spoke of Andrew Blackburn, who will serve as the Consulting Director for RNA Analytics..
development, partnerships and acquisitions will continue to shape RNA Analytics' strategic focus of developing the optimum level of answer choices underpinned with the aid of innovative technology to power client delight, long run international growth and employee and shareholder price.
This transaction combines a world-class portfolio of technology, company and individuals to generate sustained increase and force tremendous lengthy-term price. "For more than a decade, some of the most powerful possibility software solutions on present to the Actuarial world has been the Algo monetary Modeler® suite", cited Harry Kim, Board of directors, "We now have the option to work with the consumers a whole lot greater intently and supply leading edge innovation from each the technology and consulting facets. This could be carried out by now not only presenting the underlying software but also via an accelerated consulting presence throughout the regions"
Neil Collins, latest offering manager will serve because the Technical Director for RNA Analytics overseeing the product development, additional bettering and maximizing the value of the choices, in particular in two key construction areas because of the growth of Solvency II style laws and the long run implementation of IFRS17 worldwide.
The acquisition cements the tremendously integrated solution and consultancy organizations for the long-time period benefit of all present and future clients.
About RNA Analytics
RNA Analytics is a global actuarial and risk administration enterprise headquartered in Reigate, united kingdom. The RNA Analytics group is determined throughout the UK, Japan and Hong Kong, and provides actuarial and chance management consulting capabilities to economic associations world wide.
For further suggestions please contact: email@example.com
Annual global AI salary is projected to grow from $644 million in 2016 to $37 billion by way of 2025, with good use instances together with algorithmic trading strategy performance improvement; static picture cognizance, classification, and tagging; effective, scalable processing of patient statistics; predictive protection; content material distribution on social media; and greater.
The economic features trade is not any stranger to laptop discovering – a number of colossal institutions continue to effectively enforce the know-how throughout such areas as possibility analytics and law, customer segmentation, go-promoting and upselling, income and advertising and marketing crusade management, creditworthiness contrast. amongst associations that are applying computing device discovering are BBVA, JPMorgan Chase, HSBC, OCBC, and many more.
“credit applications and underwriting are the key areas the place laptop researching, and statistics analytics in usual, may have an preliminary have an impact on. The results will consist of cost rate reductions, elevated effectivity, and fewer arduous consumer experiences,” specialists suggest. McKinsey studies that in Europe, greater than a dozen banks have changed older statistical-modeling methods with machine-studying options and, in some cases, experienced 10% raises in sales of recent products, 20% discounts in capital costs, 20% raises in cash collections, and 20 percent declines in churn. The banks have carried out these features by means of devising new advice engines for clients in retailing and in small and medium-sized organizations. they have got additionally developed micro-focused fashions that greater accurately forecast who will cancel provider or default on their loans, and how most efficient to intervene, the consultancy says.
Let’s explore some exciting examples of desktop discovering purposes in banking.fiscal institutions
ML utilityBBVA Cristóbal Sepúlveda, Technical Architect at BBVA, exposed an specific use case of this technology: “At BBVA, we developed a carrier suggestion engine for bank clients. With this notion, what we are attempting to do is present the optimum business offer counting on the most used transactions through the consumer and their navigation patterns.All this advice is processed in a classification algorithm which then generates a advice. “The quantity of guidance is incredibly huge and the simplest option to present a recommendation is using laptop learning applied sciences,” he referred to. study greater on how BBVA embraces synthetic intelligence and laptop researching, in specific, Readmore here. . JPMorgan Chase At JPMorgan Chase, a gaining knowledge of machine is parsing economic deals that once saved felony teams busy for heaps of hours. The application, referred to as COIN (Contract Intelligence), does the job of decoding commercial-mortgage agreements that, except the mission wentonline in June 2016, consumed 360,000 hours of work each and every yr via legal professionals and loan officers. The utility reports files in seconds, is much less error-prone and never asks for vacation. Made viable by using investments in computer researching and a brand new private cloud network, COIN is only the start for JPMorgan Chase. The company deploy know-how hubs for groups focusing on large statistics, robotics and cloud infrastructure to discover new sources of income whereas decreasing prices and dangers. The equipment already is assisting the financial institution automate some coding actions and making its 20,000 developers greater productive, saving money.When vital, the company can additionally faucet into outdoor cloud capabilities from Amazon, Microsoft, and IBM. examine more here. HSBC The CIO at HSBC Darryl West mentioned the financial institution is the use of computing device learning to run “analytics over this huge dataset with high-quality compute means to identify patterns in the data to deliver out what looks like nefarious recreation inside our customer base. The patterns thatwe determine are then escalated to the corporations and we work with them to music down the bad guys.” The financial institution noted previous this year that it is the usage of Google Cloud computer researching capabilities for AML. read extra right here. OCBC The Singapore-primarily based OCBC financial institution has unveiled plans to use synthetic intelligence and desktop gaining knowledge of as part of its efforts to reduce monetary crimes. The financial institution intends to set up these technologies to contend with the increasing scale and complexity of AMLmonitoring, in addition to increasing the bank’s operational efficiency and accuracy within the detection of suspicious transactions.OCBC financial institution has performed a PoC with ThetaRay. Now, the business plans to birth a protracted PoC and a pre-implementation phase. The algorithm will notice anomalies in transactional conduct by way of evaluating large parameters akin to items, consumers, and risks, instead of each and every transaction as a standalone. within the PoC stage, the know-how was deployed to analyze 12 months’s value of OCBC financial institution’s company banking transaction information. The findings demonstrated that it lowered the variety of indicators, which did not require additional review, with the aid of 35%. study extra here. Lloyds Banking community Lloyds Banking group has partnered with AI startup Pindrop to make use of its machine learning expertise to realize fraudulent cell calls. Pindrop can determine 147 distinct points of a voice from a phone call or perhaps a Skype name, that can assist an individual identifyinformation such because the area that a caller is in creating an “audio fingerprint”. Lloyds Banking neighborhood will introduce the software across the Lloyds bank, Halifax and financial institution of Scotland manufacturers. Lloyds referred to the partnership with Pindrop will assist it cut down name times as well as offer protection to clients. “The cause of us doing it is about saving cash from fraud,” observed Martin Dodd, group mobilephone Managing Director at Lloyds Banking neighborhood. examine extra right here. Danske financial institution Danske financial institution, the greatest financial institution in Denmark, has created an in-condominium startup, advanced Analytics, whose sole purpose to make use of desktop researching for predictive fashions to assess client behavior and preferences on a private level. “by using inspecting client facts, wewere able to establish the consumer’s favourite skill of communique, equivalent to mobile, letter or electronic mail. [This sort of valuable info] has helped enrich our advertising campaign hit price by way of an element of 4,” says Bjørn Büchmann-Slorup, Head of AdvancedAnalytics at Danske bank. examine more here. financial institution of the united states Merrill Lynch bank of the us Merrill Lynch announced a new solution in August 2017 – clever Receivables – that makes use of synthetic intelligence and other software to help groups enrich their straight-through reconciliation (STR) of incoming payments to helpthem submit their receivables sooner. “Our solution brings collectively AI, laptop getting to know and optical persona attention (OCR), atmosphere a new bar in bills receivable reconciliation and fee matching,” added Gardner. “We’re excited to be working with main FinTech issuer HighRadius so as to add clever Receivables to our suite of options.” “financial institution of the usa Merrill Lynch’s clever Receivables answer, powered via HighRadius’ slicing-edge laptop-researching technology, will enable their company customers to accelerate the adoption of electronic funds from their end-purchasers. we're extraordinarily excited to work with BofA Merrill on modernizing treasury management capabilities and streamlining the receivables-to-cash cycle,” noted Sashi Narahari, CEO & President of HighRadius employer. study more here. Securities and exchange fee (SEC) SEC turned to advanced strategies after the 2008 crisis. “…using essential notice countsand whatever called ordinary expressions, which is a means to computer-identify structured phrases in text-based mostly documents. in a single of our first tests, we examined company company filings to verify even if we may have foreseen one of the most risks posed bythe upward thrust and use of credit default swaps [CDS] contracts main up to the financial disaster. We did this by using textual content analytic methods to computer-measure the frequency with which these contracts have been outlined in filings by using corporate issuers. We then examined thetrends across time and throughout corporate issuers to study whether any sign of impending risk emerged that could have been used as an early warning.” unless these days, SECactively reviews the potential of machines studying via continual testing throughout core activities. The economic trade Regulatory Authority (FINRA) “FINRA screens roughly 50 billion market “movements” a day, together with inventory orders, adjustments, cancellations, and trades. It looks for around 270 patterns to find potential rule violations.it would no longer say what number of activities are flagged, or how many of these yield proof of misbehavior. The desktop learning application FINRA is developing can be capable of look past those set patterns and remember which situations actually warrant purple flags.” more on how FINRA is leveraging computing device learning and artificial intelligence tocatch stock market cheaters will also be discovered read greater right here. London stock exchange (LSE) LSE has teamed up with IBM Watson enterprise and cybersecurity company SparkCognition to strengthen its AI-more suitable surveillance, stated Chris Corrado, Chief operating Officer of LSE neighborhood, in an interview with Reuters. Wells Fargo Wells Fargo analysts built a robotic referred to as AIERA (artificially clever equity analysis analyst), which is now monitoring 13 stocks. “AIERA’s fundamental aim is to song shares and formulate a regular, weekly and general view on whether the shares tracked will go up or down,” noted Ken Sena, Head of cyber web equity research. “View AIERA as improving versus changing. The months spent setting up the bot helped the team of analysts deepen their realizing of the synthetic intelligence and computing device getting to know capabilities used at most of the web groups they analyze.while AIERA is not making a choice on stocks in the traditional experience yet, her validity exams continue to indicate above general. read extra here. Goldman Sachs The financial institution has been working on a assignment dubbed “AppBank.” The initiative seeks to make use of laptop gaining knowledge of. AppBank is run via a brand new enterprise unit, which contains facts scientists and machine gaining knowledge of specialists. Its intention is to enhance “significant-scale automation” andwhile it is particularly concentrated on operations expertise, it's going to handle functions across every enterprise unit at the enterprise. “The intention is to be capable of give greater insight into the health and operations of the systems. We feel of it as our ‘investigate engine easy’ product,” observed Don Duet, head of expertise at GS. Like a light on a vehicle dashboard coming on to point out a problem, the utility would inform clients when there changed into whatever thing that might prevent the bank’s expertise infrastructure from operating easily. study greater here. . Being probably the most forward-thinking associations, Goldman Sachs has mighty ties (as a client and as an investor) with AI software company Digital Reasoning ,whose solution GS makes use of to song traders . The equal startup has additionally launched a programwith NASDAQ to use its AI expertise to track buying and selling records, communications, emails, chats and even voice facts to ferret out misconduct throughout the total digital stock alternate. Goldman Sachs additionally uses the laptop discovering platform Kensho to mine datafrom the national Bureau of Labor records and assemble all that tips into typical summaries. The stories feature 13 reveals predicting stock performances in accordance with identical employment adjustments during the past, and they’re ready to print simply 9 minutes after the information is entered.
The areas the place laptop discovering will have gigantic impact gurus emphasize chance administration; compliance; financial crime, fraud detection, and cybersecurity; credit underwriting and portfolio monitoring; consumer earnings and repair.
The Western impartial Bankers (WIB) shares that banks and FinTech businesses already use computer gaining knowledge of to discover fraud via flagging ordinary transactions. Such anomalies are investigated, with the effect being fed lower back into the gadget so it can be taught and therefore further build the client profile. The method is far more efficient than human manual monitoring and is anticipated to develop into the norm in banking and finance.
“whereas outdated fiscal fraud detection methods depended closely on advanced and potent units of guidelines, modern fraud detection goes beyond following a checklist of chance factors – it actively learns and calibrates to new expertise (or actual) safety threats. this is the vicinity of desktop studying in finance for fraud – however the identical concepts hold authentic for other facts safety problems. the usage of computer researching, techniques can realize wonderful activities or behaviors (“anomalies”) and flag them for protection teams. The problem for these techniques is to steer clear of false-positives – cases the place “dangers” are flagged that have been by no means dangers within the first place,” Tech Emergence emphasizes.
supply: Demystifying computer getting to know for Banking, Feedzai
In a tough banking environment, banks want to machine discovering to in the reduction of prices and raise retention. analysis means that banks which have changed older statistical-modeling approaches to credit risk with laptop researching strategies have experienced up to 20% raises in money collections from magnificent loans.
developments in computer discovering and big facts have pushed exchange in how systematic managers incorporate records, technology, and analytics into their funding manner. gurus imply that 62% of the systematic managers are using machine discovering concepts within the funding technique.
*Featured image credit score: Técnico Lisboa.
live clean on FinTech. Get our each day Insights.
consumer churn is a problem that each one organizations deserve to monitor, especially people that depend upon subscription-based mostly salary streams. The essential truth is that most corporations have information that will also be used to target these people and to be aware the important thing drivers of churn, and we have Keras for Deep learning attainable in R (yes, in R!!), which anticipated consumer churn with 82% accuracy. We’re super excited for this text as a result of we are using the new keras kit to produce an artificial Neural community (ANN) mannequin on the IBM Watson Telco consumer Churn facts Set! As for many company problems, it’s equally crucial to clarify what elements pressure the model, which is why we’ll use the lime kit for explainability. We cross-checked the LIME outcomes with a Correlation evaluation the use of the corrr package. We’re no longer done yet. in addition, we use three new applications to assist with desktop gaining knowledge of (ML): recipes for preprocessing, rsample for sampling facts and yardstick for mannequin metrics. These are enormously new additions to CRAN developed by means of Max Kuhn at RStudio (creator of the caret package). It appears that R is promptly developing ML equipment that rival Python. decent news if you’re drawn to applying Deep getting to know in R! we are so let’s get going!!customer Churn: Hurts revenue, Hurts business
customer churn refers to the circumstance when a customer ends their relationship with a company, and it’s a expensive problem. clients are the gasoline that powers a business. lack of shoppers affects revenue. further, it’s a great deal greater intricate and dear to benefit new purchasers than it is to preserve present shoppers. in consequence, agencies should center of attention on decreasing client churn.
The respectable news is that desktop studying can support. for a lot of companies that offer subscription based capabilities, it’s critical to each predict customer churn and clarify what points relate to consumer churn. Older ideas such as logistic regression may also be less correct than newer suggestions similar to deep gaining knowledge of, which is why we're going to exhibit you how to model an ANN in R with the keras kit.Churn Modeling With synthetic Neural Networks (Keras)
artificial Neural Networks (ANN) are now a staple inside the sub-field of computer discovering referred to as Deep learning. Deep gaining knowledge of algorithms may also be vastly sophisticated to natural regression and classification methods (e.g. linear and logistic regression) on account of the capacity to mannequin interactions between elements that might in any other case go undetected. The challenge turns into explainability, which is frequently crucial to aid the company case. The respectable news is we get the better of each worlds with keras and lime.IBM Watson Dataset (where We bought The statistics)
A telecommunications enterprise [Telco] is involved in regards to the variety of purchasers leaving their landline enterprise for cable competitors. They deserve to have in mind who's leaving. imagine that you just’re an analyst at this company and you have to find out who is leaving and why.
The dataset comprises assistance about:
in this example we display you the way to use keras to advance a complicated and tremendously correct deep researching model in R. We stroll you in the course of the preprocessing steps, investing time into how to format the facts for Keras. We investigate cross-check the quite a lot of classification metrics, and reveal that an un-tuned ANN model can effortlessly get 82% accuracy on the unseen data. here’s the deep discovering practicing history visualization.
we've some enjoyable with preprocessing the statistics (yes, preprocessing can actually be fun and straightforward!). We use the new recipes equipment to simplify the preprocessing workflow.
We conclusion via showing you the way to explain the ANN with the lime package. Neural networks was frowned upon as a result of the “black container” nature meaning these refined models (ANNs are totally accurate) are intricate to clarify the use of ordinary strategies. no longer no extra with LIME! here’s the function magnitude visualization.
We also move-checked the LIME results with a Correlation evaluation the use of the corrr package. here’s the correlation visualization.
We even developed an ML-Powered Interactive PowerBI net utility with a client Scorecard to computer screen customer churn chance and to make strategies on how to enrich consumer fitness! suppose free to take it for a spin.
View in Full screen Mode for most efficient journeycredit
We saw that simply last week the same Telco client churn dataset became used in the article, Predict consumer Churn – Logistic Regression, decision Tree and Random woodland. We concept the article changed into staggering.
this text takes a unique strategy with Keras, LIME, Correlation evaluation, and just a few different leading edge applications. We encourage the readers to check out both articles as a result of, however the problem is an identical, both solutions are beneficial to these gaining knowledge of facts science and advanced modeling.necessities
We use the following libraries during this tutorial:
install right here programs with deploy.applications().pkgs <- c("keras", "lime", "tidyquant", "rsample", "recipes", "yardstick", "corrr") installation.programs(pkgs) Load Libraries
Load the libraries.# Load libraries library(keras) library(lime) library(tidyquant) library(rsample) library(recipes) library(yardstick) library(corrr)
in case you have not previously run Keras in R, you are going to need to installation Keras using the install_keras() characteristic.# installation Keras in case you have not put in earlier than install_keras() Import data
download the IBM Watson Telco statistics Set right here. subsequent, use read_csv() to import the statistics into a nice tidy statistics frame. We use the glimpse() feature to directly inspect the information. we have the target “Churn” and all other variables are skills predictors. The raw facts set needs to be cleaned and preprocessed for ML.# Import records churn_data_raw <- read_csv("WA_Fn-UseC_-Telco-consumer-Churn.csv") glimpse(churn_data_raw) ## Observations: 7,043 ## Variables: 21 ## $ customerID "7590-VHVEG", "5575-GNVDE", "3668-QPYBK"... ## $ gender "female", "Male", "Male", "Male", "Femal... ## $ SeniorCitizen 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0... ## $ companion "yes", "No", "No", "No", "No", "No", "No... ## $ Dependents "No", "No", "No", "No", "No", "No", "yes... ## $ tenure 1, 34, 2, 45, 2, 8, 22, 10, 28, 62, 13, ... ## $ PhoneService "No", "sure", "sure", "No", "sure", "yes", ... ## $ MultipleLines "No cell provider", "No", "No", "No phon... ## $ InternetService "DSL", "DSL", "DSL", "DSL", "Fiber optic... ## $ OnlineSecurity "No", "yes", "sure", "sure", "No", "No", "... ## $ OnlineBackup "yes", "No", "sure", "No", "No", "No", "Y... ## $ DeviceProtection "No", "sure", "No", "sure", "No", "yes", "... ## $ TechSupport "No", "No", "No", "yes", "No", "No", "No... ## $ StreamingTV "No", "No", "No", "No", "No", "sure", "Ye... ## $ StreamingMovies "No", "No", "No", "No", "No", "yes", "No... ## $ Contract "Month-to-month", "365 days", "Month-to-... ## $ PaperlessBilling "sure", "No", "sure", "No", "yes", "yes", ... ## $ PaymentMethod "digital check", "Mailed check", "Mai... ## $ MonthlyCharges 29.eighty five, fifty six.ninety five, fifty three.85, 42.30, 70.70, 99.sixty five... ## $ TotalCharges 29.85, 1889.50, 108.15, 1840.seventy five, 151.sixty five,... ## $ Churn "No", "No", "yes", "No", "sure", "yes", "... Preprocess information
We’ll go through a couple of steps to preprocess the facts for ML. First, we “prune” the information, which is nothing greater than disposing of unnecessary columns and rows. Then we cut up into working towards and trying out units. After that we explore the training set to uncover transformations that may be essential for deep getting to know. We keep the premiere for ultimate. We conclusion with the aid of preprocessing the records with the new recipes kit.Prune The records
The information has a few columns and rows we’d want to eliminate:
We’ll perform the cleansing operation with one tidyverse pipe (%>%) chain.# eradicate unnecessary information churn_data_tbl <- churn_data_raw %>% select(-customerID) %>% drop_na() %>% select(Churn, every little thing()) glimpse(churn_data_tbl) ## Observations: 7,032 ## Variables: 20 ## $ Churn "No", "No", "sure", "No", "sure", "sure", "... ## $ gender "feminine", "Male", "Male", "Male", "Femal... ## $ SeniorCitizen 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0... ## $ associate "sure", "No", "No", "No", "No", "No", "No... ## $ Dependents "No", "No", "No", "No", "No", "No", "sure... ## $ tenure 1, 34, 2, forty five, 2, 8, 22, 10, 28, sixty two, 13, ... ## $ PhoneService "No", "sure", "yes", "No", "sure", "yes", ... ## $ MultipleLines "No mobilephone carrier", "No", "No", "No phon... ## $ InternetService "DSL", "DSL", "DSL", "DSL", "Fiber optic... ## $ OnlineSecurity "No", "yes", "sure", "yes", "No", "No", "... ## $ OnlineBackup "sure", "No", "sure", "No", "No", "No", "Y... ## $ DeviceProtection "No", "sure", "No", "yes", "No", "sure", "... ## $ TechSupport "No", "No", "No", "sure", "No", "No", "No... ## $ StreamingTV "No", "No", "No", "No", "No", "yes", "Ye... ## $ StreamingMovies "No", "No", "No", "No", "No", "yes", "No... ## $ Contract "Month-to-month", "one year", "Month-to-... ## $ PaperlessBilling "yes", "No", "yes", "No", "yes", "sure", ... ## $ PaymentMethod "digital examine", "Mailed determine", "Mai... ## $ MonthlyCharges 29.85, 56.95, fifty three.85, forty two.30, 70.70, 99.65... ## $ TotalCharges 29.eighty five, 1889.50, 108.15, 1840.75, 151.sixty five,... break up Into teach/look at various sets
we've a new kit, rsample, which is terribly positive for sampling strategies. It has the initial_split() feature for splitting statistics units into practising and checking out units. The return is a distinct rsplit object.# split verify/practicing sets set.seed(a hundred) train_test_split <- initial_split(churn_data_tbl, prop = 0.8) train_test_split ## <5626/1406/7032>
we can retrieve our working towards and testing sets the usage of practising() and trying out() features.# Retrieve coach and test units train_tbl <- working towards(train_test_split) test_tbl <- testing(train_test_split) Exploration: What Transformation Steps Are necessary For ML?
This phase of the analysis is often called exploratory analysis, but definitely we are trying to reply the question, “What steps are necessary to put together for ML?” the important thing thought is figuring out what transformations are essential to run the algorithm most readily. artificial Neural Networks are ideal when the information is one-hot encoded, scaled and founded. additionally, other transformations may be a good option as smartly to make relationships simpler for the algorithm to establish. A full exploratory evaluation isn't practical listed here. With that observed we’ll cover a few assistance on transformations that can help as they relate to this dataset. in the subsequent area, we will put into effect the preprocessing recommendations.Discretize The “tenure” function
Numeric points like age, years labored, length of time capable can generalize a group (or cohort). We see this in advertising and marketing plenty (believe “millennials”, which identifies a gaggle born in a definite timeframe). The “tenure” function falls into this class of numeric features that will also be discretized into businesses.
we will split into six cohorts that divide up the person base by using tenure in roughly one year (12 month) increments. This should still help the ML algorithm discover if a group is greater/less prone to customer churn.
transform The “TotalCharges” feature
What we don’t like to see is when a lot of observations are bunched inside a small part of the range.
we can use a log transformation to even out the facts into extra of a normal distribution. It’s no longer ultimate, but it’s short and simple to get our statistics opened up a little more.
professional Tip: a brief look at various is to see if the log transformation increases the magnitude of the correlation between “TotalCharges” and “Churn”. We’ll use a few dplyr operations together with the corrr equipment to function a short correlation.
The correlation between “Churn” and “LogTotalCharges” is most fulfilling in magnitude indicating the log transformation may still enhance the accuracy of the ANN model we construct. therefore, we may still function the log transformation.One-sizzling Encoding
One-hot encoding is the method of converting specific facts to sparse facts, which has columns of best zeros and ones (here's also called developing “dummy variables” or a “design matrix”). All non-numeric facts will need to be converted to dummy variables. this is standard for binary yes/No facts as a result of we are able to comfortably convert to 1’s and nil’s. It turns into a bit greater complicated with distinct categories, which requires growing new columns of 1’s and zero`s for every category (basically one much less). we have four elements that are multi-category: Contract, information superhighway provider, multiple lines, and charge method.
ANN’s customarily perform faster and sometimes times with larger accuracy when the points are scaled and/or normalized (aka centered and scaled, also called standardizing). because ANNs use gradient descent, weights tend to replace sooner. in keeping with Sebastian Raschka, an authority within the box of Deep getting to know, a couple of examples when function scaling is vital are:
The fascinated reader can examine Sebastian Raschka’s article for a full dialogue on the scaling/normalization theme. seasoned Tip: When in doubt, standardize the records.Preprocessing With Recipes
Let’s implement the preprocessing steps/transformations uncovered all over our exploration. Max Kuhn (creator of caret) has been putting some work into Rlang ML equipment lately, and the payoff is starting to take shape. a new equipment, recipes, makes growing ML information preprocessing workflows a breeze! It takes a bit getting used to, however I’ve discovered that it in reality helps manage the preprocessing steps. We’ll go over the nitty gritty because it applies to this issue.Step 1: Create A Recipe
A “recipe” is nothing more than a collection of steps you might like to function on the practicing, checking out and/or validation units. think of preprocessing records like baking a cake (I’m now not a baker but live with me). The recipe is our steps to make the cake. It doesn’t do anything apart from create the playbook for baking.
We use the recipe() function to implement our preprocessing steps. The characteristic takes a familiar object argument, which is a modeling feature reminiscent of object = Churn ~ . that means “Churn” is the result (aka response, predictor, goal) and all other elements are predictors. The function additionally takes the information argument, which gives the “recipe steps” viewpoint on a way to follow all over baking (next).
A recipe is not very beneficial unless we add “steps”, that are used to radically change the records all through baking. The package contains a couple of positive “step features” that may also be applied. The complete record of Step services may also be viewed here. For our mannequin, we use:
The last step is to put together the recipe with the prep() characteristic. This step is used to “estimate the necessary parameters from a practising set that can later be applied to other facts units”. here is important for centering and scaling and other functions that use parameters defined from the working towards set.
right here’s how essential it's to implement the preprocessing steps that we went over!# Create recipe rec_obj <- recipe(Churn ~ ., information = train_tbl) %>% step_discretize(tenure, alternatives = checklist(cuts = 6)) %>% step_log(TotalCharges) %>% step_dummy(all_nominal(), -all_outcomes()) %>% step_center(all_predictors(), -all_outcomes()) %>% step_scale(all_predictors(), -all_outcomes()) %>% prep(statistics = train_tbl) ## step 1 discretize training ## step 2 log training ## step 3 dummy working towards ## step four middle practicing ## step 5 scale training
we will print the recipe object if we ever neglect what steps had been used to put together the records. seasoned Tip: we will keep the recipe object as an RDS file the use of saveRDS(), and then use it to bake() (discussed next) future raw facts into ML-in a position facts in construction!# Print the recipe itemrec_obj ## facts Recipe ## ## Inputs: ## ## function #variables ## result 1 ## predictor 19 ## ## practicing information contained 5626 data points and no lacking records. ## ## Steps: ## ## Dummy variables from tenure [trained] ## Log transformation on TotalCharges [trained] ## Dummy variables from ~gender, ~partner, ... [trained] ## Centering for SeniorCitizen, ... [trained] ## Scaling for SeniorCitizen, ... [trained] Step 2: Baking together with your Recipe
Now for the enjoyable half! we are able to follow the “recipe” to any information set with the bake() feature, and it tactics the statistics following our recipe steps. We’ll follow to our practicing and trying out records to convert from uncooked information to a computing device discovering dataset. assess our practising set out with glimpse(). Now that’s an ML-capable dataset prepared for ANN modeling!!# Predictors x_train_tbl <- bake(rec_obj, newdata = train_tbl) x_test_tbl <- bake(rec_obj, newdata = test_tbl) glimpse(x_train_tbl) ## Observations: 5,626 ## Variables: 35 ## $ SeniorCitizen -0.4351959, -0.4351... ## $ MonthlyCharges -1.1575972, -0.2601... ## $ TotalCharges -2.275819130, 0.389... ## $ gender_Male -1.0016900, 0.99813... ## $ Partner_Yes 1.0262054, -0.97429... ## $ Dependents_Yes -0.6507747, -0.6507... ## $ tenure_bin1 2.1677790, -0.46121... ## $ tenure_bin2 -0.4389453, -0.4389... ## $ tenure_bin3 -0.4481273, -0.4481... ## $ tenure_bin4 -0.4509837, 2.21698... ## $ tenure_bin5 -0.4498419, -0.4498... ## $ tenure_bin6 -0.4337508, -0.4337... ## $ PhoneService_Yes -3.0407367, 0.32880... ## $ MultipleLines_No.cell.provider 3.0407367, -0.32880... ## $ MultipleLines_Yes -0.8571364, -0.8571... ## $ InternetService_Fiber.optic -0.8884255, -0.8884... ## $ InternetService_No -0.5272627, -0.5272... ## $ OnlineSecurity_No.cyber web.provider -0.5272627, -0.5272... ## $ OnlineSecurity_Yes -0.6369654, 1.56966... ## $ OnlineBackup_No.internet.provider -0.5272627, -0.5272... ## $ OnlineBackup_Yes 1.3771987, -0.72598... ## $ DeviceProtection_No.internet.service -0.5272627, -0.5272... ## $ DeviceProtection_Yes -0.7259826, 1.37719... ## $ TechSupport_No.internet.provider -0.5272627, -0.5272... ## $ TechSupport_Yes -0.6358628, -0.6358... ## $ StreamingTV_No.internet.provider -0.5272627, -0.5272... ## $ StreamingTV_Yes -0.7917326, -0.7917... ## $ StreamingMovies_No.internet.provider -0.5272627, -0.5272... ## $ StreamingMovies_Yes -0.797388, -0.79738... ## $ Contract_One.year -0.5156834, 1.93882... ## $ Contract_Two.year -0.5618358, -0.5618... ## $ PaperlessBilling_Yes 0.8330334, -1.20021... ## $ PaymentMethod_Credit.card..automated. -0.5231315, -0.5231... ## $ PaymentMethod_Electronic.investigate 1.4154085, -0.70638... ## $ PaymentMethod_Mailed.investigate -0.5517013, 1.81225... Step three: Don’t neglect The target
One last step, we should keep the specific values (truth) as y_train_vec and y_test_vec, which are vital for modeling our ANN. We convert to a collection of numeric ones and zeros which may also be permitted through the Keras ANN modeling functions. We add “vec” to the identify so that you can without difficulty be aware the class of the object (it’s easy to get puzzled when working with tibbles, vectors, and matrix facts varieties).# Response variables for working towards and trying out sets y_train_vec <- ifelse(pull(train_tbl, Churn) == "yes", 1, 0) y_test_vec <- ifelse(pull(test_tbl, Churn) == "sure", 1, 0) model customer Churn With Keras (Deep learning)
here's tremendous exciting!! eventually, Deep researching with Keras in R! The group at RStudio has carried out surprising work currently to create the keras kit, which implements Keras in R. Very cool!historical past On artifical Neural Networks
For these unfamiliar with Neural Networks (and those that need a refresher), read this article. It’s very complete, and also you’ll go away with a established realizing of the kinds of deep learning and the way they work.
supply: Xenon Stack
Deep researching has been accessible in R for some time, however the primary applications used within the wild haven't (this comprises Keras, Tensor movement, Theano, etc, which might be all Python libraries). It’s value bringing up that a couple of other Deep researching programs exist in R including h2o, mxnet, and others. The fascinated reader can check out this weblog submit for a comparison of deep discovering applications in R.constructing A Deep discovering model
We’re going to build a unique classification of ANN called a Multi-Layer Perceptron (MLP). MLPs are probably the most easiest sorts of deep gaining knowledge of, however they are each enormously correct and serve as a leaping-off point for extra complicated algorithms. MLPs are fairly versatile as they can be used for regression, binary and multi classification (and are customarily reasonably respectable at classification problems).
We’ll construct a three layer MLP with Keras. Let’s stroll-through the steps earlier than we enforce in R.
Initialize a sequential model: the 1st step is to initialize a sequential mannequin with keras_model_sequential(), which is the starting of our Keras model. The sequential mannequin is composed of a linear stack of layers.
apply layers to the sequential mannequin: Layers include the enter layer, hidden layers and an output layer. The enter layer is the information and supplied it’s formatted appropriately there’s nothing more to talk about. The hidden layers and output layers are what controls the ANN internal workings.
Hidden Layers: Hidden layers form the neural community nodes that allow non-linear activation the use of weights. The hidden layers are created using layer_dense(). We’ll add two hidden layers. We’ll observe devices = sixteen, which is the variety of nodes. We’ll choose kernel_initializer = "uniform" and activation = "relu" for both layers. the primary layer needs to have the input_shape = 35, which is the number of columns within the practising set. Key element: while we are arbitrarily identifying the number of hidden layers, gadgets, kernel initializers and activation functions, these parameters will also be optimized via a system known as hyperparameter tuning it's discussed in next Steps.
Dropout Layers: Dropout layers are used to control overfitting. This eliminates weights under a cutoff threshold to steer clear of low weights from overfitting the layers. We use the layer_dropout() function add two drop out layers with cost = 0.10 to eradicate weights under 10%.
Output Layer: The output layer specifies the form of the output and the method of assimilating the realized suggestions. The output layer is applied the usage of the layer_dense(). For binary values, the form may still be gadgets = 1. For multi-classification, the devices should still correspond to the variety of courses. We set the kernel_initializer = "uniform" and the activation = "sigmoid" (common for binary classification).
assemble the model: The final step is to collect the model with bring together(). We’ll use optimizer = "adam", which is among the most widespread optimization algorithms. We choose loss = "binary_crossentropy" seeing that here is a binary classification issue. We’ll opt for metrics = c("accuracy") to be evaluated all through practising and trying out. Key aspect: The optimizer is regularly covered in the tuning process.
Let’s codify the discussion above to construct our Keras MLP-flavored ANN model.# constructing our synthetic Neural community model_keras <- keras_model_sequential() model_keras %>% # First hidden layer layer_dense( contraptions = sixteen, kernel_initializer = "uniform", activation = "relu", input_shape = ncol(x_train_tbl)) %>% # Dropout to steer clear of overfitting layer_dropout(rate = 0.1) %>% # second hidden layer layer_dense( contraptions = 16, kernel_initializer = "uniform", activation = "relu") %>% # Dropout to stay away from overfitting layer_dropout(cost = 0.1) %>% # Output layer layer_dense( devices = 1, kernel_initializer = "uniform", activation = "sigmoid") %>% # compile ANN compile( optimizer = 'adam', loss = 'binary_crossentropy', metrics = c('accuracy') ) model_keras ## model ## ______________________________________________________________________ ## Layer (type) Output form Param # ## ====================================================================== ## dense_1 (Dense) (None, 16) 576 ## ______________________________________________________________________ ## dropout_1 (Dropout) (None, sixteen) 0 ## ______________________________________________________________________ ## dense_2 (Dense) (None, sixteen) 272 ## ______________________________________________________________________ ## dropout_2 (Dropout) (None, sixteen) 0 ## ______________________________________________________________________ ## dense_3 (Dense) (None, 1) 17 ## ====================================================================== ## total params: 865 ## Trainable params: 865 ## Non-trainable params: 0 ## ______________________________________________________________________
We use the healthy() feature to run the ANN on our practising information. the item is our mannequin, and x and y are our practising information in matrix and numeric vector kinds, respectively. The batch_size = 50 sets the quantity samples per gradient update inside each epoch. We set epochs = 35 to manage the number practising cycles. customarily we wish to maintain the batch measurement excessive seeing that this decreases the error inside each and every practising cycle (epoch). We additionally need epochs to be significant, which is essential in visualizing the practising historical past (discussed under). We set validation_split = 0.30 to include 30% of the records for model validation, which prevents overfitting. The training procedure should still complete in 15 seconds or so.# healthy the keras model to the practising records fit_keras <- healthy( object = model_keras, x = as.matrix(x_train_tbl), y = y_train_vec, batch_size = 50, epochs = 35, validation_split = 0.30 )
we can inspect the ultimate mannequin. We are looking to be sure there is minimal difference between the validation accuracy and the practising accuracy.# Print the ultimate modelfit_keras ## expert on 3,938 samples, validated on 1,688 samples (batch_size=50, epochs=35) ## last epoch (plot to look heritage): ## val_loss: 0.4215 ## val_acc: 0.8057 ## loss: 0.399 ## acc: 0.8101
we can visualize the Keras working towards history the usage of the plot() function. What we need to see is the validation accuracy and loss leveling off, which potential the model has completed working towards. We see that there is some divergence between training loss/accuracy and validation loss/accuracy. This mannequin shows we will perhaps stop practicing at an past epoch. pro Tip: most effective use enough epochs to get a high validation accuracy. once validation accuracy curve starts off to flatten or decrease, it’s time to stop working towards.# Plot the training/validation background of our Keras brandplot(fit_keras) + theme_tq() + scale_color_tq() + scale_fill_tq() + labs(title = "Deep studying practicing effects")
We’ve received an excellent model in line with the validation accuracy. Now let’s make some predictions from our keras model on the examine data set, which was unseen right through modeling (we use this for the authentic performance assessment). we have two functions to generate predictions:
The yardstick kit has a set of convenient capabilities for measuring efficiency of laptop learning fashions. We’ll overview some metrics we can use to understand the performance of our mannequin.
First, let’s get the information formatted for yardstick. We create an information frame with the fact (exact values as components), estimate (predicted values as elements), and the category likelihood (chance of yes as numeric). We use the fct_recode() characteristic from the forcats kit to aid with recoding as yes/No values.# layout verify data and predictions for yardstick metrics estimates_keras_tbl <- tibble( reality = as.component(y_test_vec) %>% fct_recode(sure = "1", no = "0"), estimate = as.aspect(yhat_keras_class_vec) %>% fct_recode(sure = "1", no = "0"), class_prob = yhat_keras_prob_vec ) estimates_keras_tbl ## # A tibble: 1,406 x 3 ## certainty estimate class_prob ## ## 1 yes no 0.328355074 ## 2 yes yes 0.633630514 ## three no no 0.004589651 ## 4 no no 0.007402068 ## 5 no no 0.049968336 ## 6 no no 0.116824441 ## 7 no sure 0.775479317 ## 8 no no 0.492996633 ## 9 no no 0.011550998 ## 10 no no 0.004276015 ## # ... with 1,396 extra rows
Now that we've the facts formatted, we will take abilities of the yardstick kit.Confusion desk
we will use the conf_mat() feature to get the confusion desk. We see that the model turned into by using no ability best, nevertheless it did an honest job of picking out valued clientele prone to churn.# Confusion table estimates_keras_tbl %>% conf_mat(actuality, estimate) ## fact ## Prediction no sure ## no 950 161 ## sure 99 196 Accuracy
we can use the metrics() characteristic to get an accuracy measurement from the check set. We are getting roughly eighty two% accuracy.# Accuracy estimates_keras_tbl %>% metrics(reality, estimate) ## # A tibble: 1 x 1 ## accuracy ## ## 1 0.8150782 AUC
we will also get the ROC enviornment under the Curve (AUC) dimension. AUC is regularly a pretty good metric used to compare distinct classifiers and to evaluate to randomly guessing (AUC_random = 0.50). Our mannequin has AUC = 0.eighty five, which is an awful lot more advantageous than randomly guessing. Tuning and checking out different classification algorithms may additionally yield even superior outcomes.# AUC estimates_keras_tbl %>% roc_auc(certainty, class_prob) ##  0.8523951 Precision And bear in mind
Precision is when the mannequin predicts “yes”, how regularly is it basically “sure”. keep in mind (additionally authentic tremendous cost or specificity) is when the genuine price is “yes” how regularly is the mannequin appropriate. we will get precision() and keep in mind() measurements using yardstick.# Precision tibble( precision = estimates_keras_tbl %>% precision(actuality, estimate), take into account = estimates_keras_tbl %>% bear in mind(certainty, estimate) ) ## # A tibble: 1 x 2 ## precision consider ## ## 1 0.8550855 0.9056244
Precision and take into account are very crucial to the business case: The firm is worried with balancing the can charge of concentrated on and holding customers vulnerable to leaving with the can charge of inadvertently targeting valued clientele that are not planning to go away (and probably reducing income from this community). the brink above which to foretell Churn = “sure” will also be adjusted to optimize for the business issue. This turns into an consumer Lifetime price optimization problem it is mentioned additional in subsequent Steps.F1 rating
we can additionally get the F1-ranking, which is a weighted general between the precision and don't forget. machine learning classifier thresholds are often adjusted to maximize the F1-ranking. although, here is frequently now not the most reliable solution to the business issue.# F1-Statistic estimates_keras_tbl %>% f_meas(truth, estimate, beta = 1) ##  0.8796296 clarify The mannequin With LIME
LIME stands for native Interpretable mannequin-agnostic Explanations, and is a method for explaining black-container machine learning mannequin classifiers. For those new to LIME, this YouTube video does a very first-rate job explaining how LIME helps to determine characteristic value with black container computer researching models (e.g. deep researching, stacked ensembles, random woodland).Setup
The lime equipment implements LIME in R. One issue to note is that it’s no longer setup out-of-the-field to work with keras. The good news is with a couple of services we can get every little thing working adequately. We’ll need to make two customized features:
model_type: Used to tell lime what classification of mannequin we are coping with. It could be classification, regression, survival, and many others.
predict_model: Used to permit lime to operate predictions that its algorithm can interpret.
the first factor we should do is determine the class of our model object. We try this with the category() function.category(model_keras) ##  "keras.models.Sequential" ##  "keras.engine.practising.model" ##  "keras.engine.topology.Container" ##  "keras.engine.topology.Layer" ##  "python.builtin.object"
subsequent we create our model_type() function. It’s best input is x the keras mannequin. The characteristic simply returns “classification”, which tells LIME we're classifying.# Setup lime::model_type() function for keras model_type.keras.fashions.Sequential <- function(x, ...) return("classification")
Now we are able to create our predict_model() feature, which wraps keras::predict_proba(). The trick here is to understand that it’s inputs need to be x a mannequin, newdata a dataframe object (here's essential), and kind which is not used however can also be use to swap the output classification. The output is additionally a bit elaborate because it have to be in the format of percentages with the aid of classification (this is critical; shown subsequent).# Setup lime::predict_model() feature for keras predict_model.keras.fashions.Sequential <- function(x, newdata, classification, ...) pred <- predict_proba(object = x, x = as.matrix(newdata)) return(statistics.frame(yes = pred, No = 1 - pred))
Run this next script to display you what the output looks like and to test our predict_model() feature. See how it’s the probabilities by classification. It ought to be during this form for model_type = "classification".# check our predict_model() function predict_model(x = model_keras, newdata = x_test_tbl, category = 'raw') %>% tibble::as_tibble() ## # A tibble: 1,406 x 2 ## sure No ## ## 1 0.328355074 0.6716449 ## 2 0.633630514 0.3663695 ## 3 0.004589651 0.9954103 ## four 0.007402068 0.9925979 ## 5 0.049968336 0.9500317 ## 6 0.116824441 0.8831756 ## 7 0.775479317 0.2245207 ## 8 0.492996633 0.5070034 ## 9 0.011550998 0.9884490 ## 10 0.004276015 0.9957240 ## # ... with 1,396 more rows
Now the enjoyable half, we create an explainer the use of the lime() characteristic. simply circulate the training facts set devoid of the “Attribution column”. The form have to be an information frame, which is adequate given that our predict_model characteristic will switch it to an keras object. Set model = automl_leader our leader mannequin, and bin_continuous = FALSE. We might tell the algorithm to bin continuous variables, but this may no longer make experience for express numeric statistics that we didn’t change to components.# Run lime() on practising set explainer <- lime::lime( x = x_train_tbl, mannequin = model_keras, bin_continuous = FALSE)
Now we run the clarify() characteristic, which returns our rationalization. this may take a minute to run so we limit it to just the primary ten rows of the test records set. We set n_labels = 1 because we care about explaining a single class. surroundings n_features = 4 returns the right four facets which are important to each case. at last, surroundings kernel_width = 0.5 permits us to boost the “model_r2” value via shrinking the localized comparison.# Run clarify() on explainer clarification <- lime::explain( x_test_tbl[1:10,], explainer = explainer, n_labels = 1, n_features = 4, kernel_width = 0.5) function magnitude Visualization
The payoff for the work we put in the use of LIME is that this function magnitude plot. This allows us to visualize each and every of the first ten circumstances (observations) from the check information. The proper 4 aspects for every case are shown. word that they are not the equal for each and every case. The green bars imply that the function helps the mannequin conclusion, and the crimson bars contradict. a number of critical aspects in accordance with frequency in first ten circumstances:
another amazing visualization may also be performed the use of plot_explanations(), which produces a facetted heatmap of all case/label/feature combinations. It’s a extra condensed version of plot_features(), but we need to be cautious because it does not give accurate data and it makes it less easy to investigate binned aspects (observe that “tenure” would not be identified as a contributor besides the fact that it indicates up as a true feature in 7 of 10 situations).plot_explanations(rationalization) + labs(title = "LIME feature magnitude Heatmap", subtitle = "dangle Out (verify) Set, First 10 cases shown")
investigate Explanations With Correlation evaluation
One element we deserve to be careful with the LIME visualization is that we're best doing a sample of the facts, in our case the primary 10 verify observations. hence, we're gaining a very localized knowing of how the ANN works. youngsters, we also want to be aware of on from a worldwide viewpoint what drives feature significance.
we will operate a correlation evaluation on the practising set as neatly to help glean what points correlate globally to “Churn”. We’ll use the corrr package, which performs tidy correlations with the function correlate(). we are able to get the correlations as follows.# feature correlations to Churn corrr_analysis <- x_train_tbl %>% mutate(Churn = y_train_vec) %>% correlate() %>% focus(Churn) %>% rename(feature = rowname) %>% organize(abs(Churn)) %>% mutate(function = as_factor(characteristic)) corrr_analysis ## # A tibble: 35 x 2 ## function Churn ## ## 1 gender_Male -0.006690899 ## 2 tenure_bin3 -0.009557165 ## three MultipleLines_No.mobilephone.service -0.016950072 ## four PhoneService_Yes 0.016950072 ## 5 MultipleLines_Yes 0.032103354 ## 6 StreamingTV_Yes 0.066192594 ## 7 StreamingMovies_Yes 0.067643871 ## eight DeviceProtection_Yes -0.073301197 ## 9 tenure_bin4 -0.073371838 ## 10 PaymentMethod_Mailed.assess -0.080451164 ## # ... with 25 greater rows
The correlation visualization helps in distinguishing which aspects are relavant to Churn.# Correlation visualization corrr_analysis %>% ggplot(aes(x = Churn, y = fct_reorder(function, desc(Churn)))) + geom_point() + # tremendous Correlations - make contributions to churn geom_segment(aes(xend = 0, yend = feature), color = palette_light()[], facts = corrr_analysis %>% filter(Churn > 0)) + geom_point(color = palette_light()[], statistics = corrr_analysis %>% filter(Churn > 0)) + # terrible Correlations - evade churn geom_segment(aes(xend = 0, yend = function), colour = palette_light()[], statistics = corrr_analysis %>% filter(Churn <</span> 0)) + geom_point(colour = palette_light()[], statistics = corrr_analysis %>% filter(Churn <</span> 0)) + # Vertical lines geom_vline(xintercept = 0, color = palette_light()[], dimension = 1, linetype = 2) + geom_vline(xintercept = -0.25, color = palette_light()[], dimension = 1, linetype = 2) + geom_vline(xintercept = 0.25, colour = palette_light()[], measurement = 1, linetype = 2) + # Aesthetics theme_tq() + labs(title = "Churn Correlation evaluation", subtitle = "fantastic Correlations (make contributions to churn), poor Correlations (prevent churn)", y = "function significance")
The correlation analysis helps us immediately disseminate which elements that the LIME evaluation could be except for. we are able to see that the following features are enormously correlated (magnitude > 0.25):
increases chance of Churn (pink):
Decreases probability of Churn (Blue):
we are able to investigate elements that are most regular in the LIME feature significance visualization together with people that the correlation analysis shows an above ordinary magnitude. We’ll investigate:
LIME instances point out that the ANN mannequin is using this function often and high correlation consents that this is essential. Investigating the function distribution, it seems that valued clientele with lower tenure (bin 1) are more likely to depart. chance: target valued clientele with lower than 12 month tenure.
Contract (totally Correlated)
while LIME didn't indicate this as a chief characteristic within the first 10 instances, the feature is evidently correlated with those electing to live. customers with one and two 12 months contracts are a whole lot less prone to churn. opportunity: present promoting to swap to long term contracts.
information superhighway carrier (totally Correlated)
while LIME didn't indicate this as a first-rate feature within the first 10 instances, the function is evidently correlated with those electing to dwell. consumers with fiber optic carrier are more likely to churn while these with out a internet service are less likely to churn. improvement enviornment: consumers can be disillusioned with fiber optic service.
payment components (totally Correlated)
while LIME did not point out this as a chief characteristic within the first 10 cases, the feature is obviously correlated with those electing to reside. clients with digital examine are more likely to leave. opportunity: offer consumers a merchandising to swap to computerized payments.
Senior Citizen (5/10 LIME instances)
Senior citizen looked in a couple of of the LIME cases indicating it was essential to the ANN for the ten samples. youngsters, it became no longer incredibly correlated to Churn, which can also point out that the ANN is the usage of in an greater subtle manner (e.g. as an interaction). It’s complicated to claim that senior citizens usually tend to leave, but non-senior citizens seem much less prone to churning. chance: goal users within the lower age demographic.
on-line protection (4/10 LIME situations)
clients that did not sign up for on-line protection were more likely to leave while valued clientele with no information superhighway service or on-line protection were less more likely to leave. opportunity: Promote on-line protection and different packages that increase retention fees.
subsequent Steps: company Science university
We’ve simply scratched the surface with the answer to this issue, but sadly there’s simplest so a great deal ground we are able to cover in a piece of writing. listed here are a few next steps that I’m joyful to announce should be coated in a business Science college route coming in 2018!client Lifetime value
Your firm must see the financial benefit so at all times tie your evaluation to sales, profitability or ROI. client Lifetime value (CLV) is a technique that ties the business profitability to the retention rate. while we did not enforce the CLV methodology herein, a full client churn evaluation would tie the churn to an classification cutoff (threshold) optimization to maximize the CLV with the predictive ANN mannequin.
The simplified CLV mannequin is:
The ANN model we built is good, but it surely may well be superior. How we have in mind our mannequin accuracy and enrich on it is in the course of the combination of two strategies:
We should put in force k-Fold go Validation and Hyper Parameter Tuning if we desire a most useful-in-type model.Distributing Analytics
It’s important to talk statistics science insights to choice makers in the corporation. Most choice makers in businesses don't seem to be statistics scientists, but these individuals make essential selections on a everyday groundwork. The PowerBI software beneath contains a consumer Scorecard to computer screen customer health (chance of churn). The utility walks the person during the laptop gaining knowledge of journey for a way the mannequin became developed, what it skill to stakeholders, and the way it can also be used in construction.
View in Full screen Mode for top-rated experience
For those in the hunt for alternatives for distributing analytics, two first rate options are:
bright Apps for speedy prototyping: brilliant net applications offer the highest flexibility with R algorithms inbuilt. vivid is greater complicated to learn, but bright applications are staggering / limitless.
Microsoft PowerBI and Tableau for Visualization: permit disbursed analytics with the capabilities of intuitive constitution however with some flexibilty sacrificed. can be complicated to build ML into.
You’re probably wondering why we're going into so a whole lot detail on next steps. we are satisfied to announce a brand new assignment for 2018: company Science school, an online school committed to helping records science newcomers enrich within the areas of:
researching paths can be focused on enterprise and economic functions. We’ll maintain you posted by the use of social media and our blog (please comply with us / subscribe to stay up to date).
Please let us know in case you are interested in becoming a member of business Science institution. let us know what you feel within the Disqus feedback below.Conclusions
client churn is a costly difficulty. The first rate information is that desktop getting to know can remedy churn problems, making the firm extra ecocnomic in the manner. listed here, we noticed how Deep learning may also be used to foretell consumer churn. We constructed an ANN model the use of the new keras equipment that carried out 82% predictive accuracy (with out tuning)! We used three new desktop learning applications to aid with preprocessing and measuring performance: recipes, rsample and yardstick. at last we used lime to explain the Deep learning model, which historically turned into unimaginable! We checked the LIME results with a Correlation analysis, which dropped at light other points to investigate. For the IBM Telco dataset, tenure, contract type, cyber web carrier category, charge menthod, senior citizen repute, and on-line protection reputation were beneficial in diagnosing client churn. We hope you enjoyed this text!About business Science
business Science specializes in “ROI-pushed statistics science”. Our focus is machine getting to know and statistics science in company and monetary functions. We assist corporations that are seeking so as to add this aggressive knowledge however can also not have the resources at present to put into effect predictive analytics. enterprise Science can help to expand into predictive analytics whereas executing on ROI producing initiatives. consult with the company Science site or contact us to gain knowledge of greater!
connectedR-bloggers.com presents daily electronic mail updates about R information and tutorials on themes reminiscent of: information science, massive information, R jobs, visualization (ggplot2, Boxplots, maps, animation), programming (RStudio, Sweave, LaTeX, SQL, Eclipse, git, hadoop, net Scraping) records (regression, PCA, time collection, buying and selling) and extra... if you got this a ways, why not subscribe for updates from the website? choose your flavor: electronic mail, twitter, RSS, or fb...
Unquestionably it is hard assignment to pick dependable certification questions/answers assets regarding review, reputation and validity since individuals get sham because of picking incorrectly benefit. Killexams.com ensure to serve its customers best to its assets concerning exam dumps update and validity. The vast majority of other's sham report dissension customers come to us for the brain dumps and pass their exams joyfully and effortlessly. We never trade off on our review, reputation and quality on the grounds that killexams review, killexams reputation and killexams customer certainty is imperative to us. Uniquely we deal with killexams.com review, killexams.com reputation, killexams.com sham report objection, killexams.com trust, killexams.com validity, killexams.com report and killexams.com scam. On the off chance that you see any false report posted by our rivals with the name killexams sham report grievance web, killexams.com sham report, killexams.com scam, killexams.com protest or something like this, simply remember there are constantly awful individuals harming reputation of good administrations because of their advantages. There are a huge number of fulfilled clients that pass their exams utilizing killexams.com brain dumps, killexams PDF questions, killexams hone questions, killexams exam simulator. Visit Killexams.com, our specimen questions and test brain dumps, our exam simulator and you will realize that killexams.com is the best brain dumps site.
Killexams C2010-504 test questions | Killexams HP0-S25 practice questions | Killexams A00-240 practical test | Killexams 000-529 braindump | Killexams HC-611 sample questions | Killexams HP0-S18 Practice Test | Killexams HP5-K03D sample test | Killexams 000-581 exam dumps | Killexams 250-421 test prep | Killexams 77-600 past exams | Killexams 1Z1-574 exam questions | Killexams 77-602 Q&A | Killexams 190-722 real test | Killexams BE-100W practice questions | Killexams HP0-X01 brain dump | Killexams PEGACBA001 real questions | Killexams VCS-254 | Killexams EE0-512 | Killexams 000-115 | Killexams 70-549-CSharp |
Pass4sure C2020-002 Dumps and Practice Tests with Real Questions
We are for the most part very much aware that a noteworthy issue in the IT business is that there is an absence of value ponder materials. Our exam readiness material gives you all that you should take a confirmation examination. Our IBM C2020-002 Exam will give you exam inquiries with confirmed answers that mirror the real exam. High caliber and incentive for the C2020-002 Exam. We at killexams.com are resolved to enable you to clear your C2020-002 accreditation test with high scores.
If you are interested in successfully completing the IBM C2020-002 Certification to start earning? killexams.com has leading edge developed IBM exam questions that will ensure you pass this C2020-002 exam! killexams.com delivers you the most accurate, current and latest updated C2020-002 Certification exam questions and available with a 100% money back guarantee promise. There are many companies that provide C2020-002 brain dumps but those are not accurate and latest ones. Preparation with killexams.com C2020-002 new questions is a best way to pass this certification exam in easy way.
We are all well aware that a major problem in the IT industry is that there is a lack of quality study materials. Our exam preparation material provides you everything you will need to take a certification examination. Our IBM C2020-002 Exam will provide you with exam questions with verified answers that reflect the actual exam. These questions and answers provide you with the experience of taking the actual test. High quality and value for the C2020-002 Exam. 100% guarantee to pass your IBM C2020-002 exam and get your IBM certification. We at killexams.com are committed to help you clear your C2020-002 certification test with high scores. The chances of you failing to clear your C2020-002 test, after going through our comprehensive exam dumps are very little.
IBM C2020-002 is ubiquitous all around the globe, and the business and programming arrangements gave by them are being grasped by every one of the organizations. They have helped in driving a large number of organizations on the beyond any doubt shot way of achievement. Far reaching learning of IBM items are viewed as a critical capability, and the experts confirmed by them are exceptionally esteemed in all associations.
We give genuine C2020-002 pdf exam inquiries and answers braindumps in two arrangements. Download PDF and Practice Tests. Pass IBM C2020-002 book Exam rapidly and effectively. The C2020-002 syllabus PDF sort is accessible for perusing and printing. You can print increasingly and rehearse ordinarily. Our pass rate is high to 98.9% and the comparability rate between our C2020-002 syllabus think about guide and genuine exam is 90% in light of our seven-year teaching background. Do you need accomplishments in the C2020-002 exam in only one attempt? I am right now examining for the IBM C2020-002 syllabus exam.
Cause the only thing that is in any way important here is passing the IBM C2020-002 exam. Cause all that you require is a high score of IBM C2020-002 exam. The just a single thing you have to do is downloading Examcollection C2020-002 exam consider directs now. We won't let you down with our unconditional promise. The experts likewise keep pace with the most up and coming exam so as to give the greater part of refreshed materials. One year free access to have the capacity to them through the date of purchase. Each applicant may bear the cost of the IBM exam dumps through killexams.com at a low cost. Frequently there is a markdown for anybody all.
Killexams.com Huge Discount Coupons and Promo Codes are as under;
WC2017 : 60% Discount Coupon for all exams on website
PROF17 : 10% Discount Coupon for Orders greater than $69
DEAL17 : 15% Discount Coupon for Orders greater than $99
DECSPECIAL : 10% Special Discount Coupon for All Orders
At killexams.com, we give totally studied IBM C2020-002 getting ready resources which are the best to clear C2020-002 test, and to get asserted by IBM. It is a best choice to animate your employment as a specialist in the Information Technology industry. We are satisfied with our reputation of helping people clear the C2020-002 test in their first undertakings. Our thriving rates in the past two years have been totally extraordinary, as a result of our cheery customers who are presently prepared to induce their livelihoods in the most optimized plan of attack. killexams.com is the primary choice among IT specialists, especially the ones who are planning to climb the movement levels faster in their individual affiliations. IBM is the business pioneer in information development, and getting avowed by them is a guaranteed way to deal with win with IT jobs. We empower you to do correctly that with our brilliant IBM C2020-002 getting ready materials.
IBM C2020-002 is omnipresent all around the world, and the business and programming courses of action gave by them are being gotten a handle on by each one of the associations. They have helped in driving an extensive number of associations on the without question shot method for accomplishment. Expansive learning of IBM things are seen as a basic capacity, and the specialists affirmed by them are uncommonly regarded in all affiliations.
We give honest to goodness C2020-002 pdf exam request and answers braindumps in two plans. Download PDF and Practice Tests. Pass IBM C2020-002 book Exam quickly and viably. The C2020-002 syllabus PDF sort is available for examining and printing. You can print progressively and practice usually. Our pass rate is high to 98.9% and the similarity rate between our C2020-002 syllabus consider manage and certifiable exam is 90% in light of our seven-year instructing foundation. Do you require achievements in the C2020-002 exam in just a single endeavor? I am at the present time analyzing for the IBM C2020-002 syllabus exam.
Cause the main thing that is in any capacity critical here is passing the IBM C2020-002 exam. Cause all that you require is a high score of IBM C2020-002 exam. The only a solitary thing you need to do is downloading Examcollection C2020-002 exam consider coordinates now. We won't let you down with our unrestricted guarantee. The specialists in like manner keep pace with the most best in class exam to give most of invigorated materials. One year free access to have the ability to them through the date of procurement. Every candidate may bear the cost of the IBM exam dumps through killexams.com requiring little to no effort. Habitually there is a markdown for anyone all.
Inside seeing the bona fide exam substance of the mind dumps at killexams.com you can without a lot of an extend develop your claim to fame. For the IT specialists, it is basic to enhance their capacities as showed by their work need. We make it basic for our customers to carry certification exam with the help of killexams.com affirmed and honest to goodness exam material. For an awesome future in its domain, our mind dumps are the best decision.
A best dumps creating is a basic segment that makes it straightforward for you to take IBM accreditations. In any case, IBM braindumps PDF offers settlement for candidates. The IT assertion is a critical troublesome endeavor if one doesn't find genuine course as obvious resource material. Thus, we have genuine and invigorated substance for the arranging of affirmation exam.
It is fundamental to collect to the guide material in case one needs toward save time. As you require packs of time to look for revived and genuine examination material for taking the IT accreditation exam. If you find that at one place, what could be better than this? It's simply killexams.com that has what you require. You can save time and maintain a strategic distance from trouble in case you buy Adobe IT accreditation from our site.
You should get the most revived IBM C2020-002 Braindumps with the correct answers, which are set up by killexams.com specialists, empowering the likelihood to understand finding out about their C2020-002 affirmation course in the best, you won't find C2020-002 consequences of such quality wherever in the market. Our IBM C2020-002 Practice Dumps are given to candidates at performing 100% in their exam. Our IBM C2020-002 test dumps are latest in the market, enabling you to prepare for your C2020-002 exam in the right way.
If you are possessed with viably completing the IBM C2020-002 Certification to start obtaining? killexams.com has driving edge made IBM exam tends to that will promise you pass this C2020-002 exam! killexams.com passes on you the most correct, present and latest revived C2020-002 Certification exam questions and open with a 100% genuine guarantee ensure. There are numerous associations that give C2020-002 mind dumps yet those are not exact and latest ones. Course of action with killexams.com C2020-002 new request is a most perfect way to deal with pass this certification exam in basic way.
Killexams.com Huge Discount Coupons and Promo Codes are as under;
WC2017 : 60% Discount Coupon for all exams on website
PROF17 : 10% Discount Coupon for Orders greater than $69
DEAL17 : 15% Discount Coupon for Orders greater than $99
DECSPECIAL : 10% Special Discount Coupon for All Orders
We are generally especially mindful that an imperative issue in the IT business is that there is a nonattendance of significant worth consider materials. Our exam preparation material gives all of you that you should take an affirmation examination. Our IBM C2020-002 Exam will give you exam request with affirmed answers that mirror the genuine exam. These request and answers give you the experience of taking the honest to goodness test. High bore and impetus for the C2020-002 Exam. 100% confirmation to pass your IBM C2020-002 exam and get your IBM attestation. We at killexams.com are made plans to empower you to clear your C2020-002 accreditation test with high scores. The chances of you fail to clear your C2020-002 test, in the wake of encountering our sweeping exam dumps are for all intents and purposes nothing.
Killexams 00M-228 exam dumps | Killexams 6207-1 test questions | Killexams C2180-276 braindump | Killexams 000-730 practice questions | Killexams COG-180 sample test | Killexams 000-914 Q&A | Killexams SSCP real questions | Killexams HP0-310 practice questions | Killexams HP0-Y36 real test | Killexams RTRP brain dump | Killexams 1Z0-863 exam questions | Killexams 190-755 sample questions | Killexams FN0-202 past exams | Killexams 700-702 test prep | Killexams A2010-654 Practice Test | Killexams CAT-140 practical test | Killexams HP2-N43 | Killexams JN0-332 | Killexams PMP | Killexams SQ0-101 |
Great opportunity to get certified C2020-002 exam.
Hearty thanks to killexams.com crew for the query & answer of C2020-002 exam. It furnished exquisite method to my queries on C2020-002 I felt confident to stand the take a look at. observed many questions inside the exam paper much like the guide. I strongly experience that the guide is still valid. respect the effort with the aid of your crew contributors, killexams.com. The method of dealing subjects in a unique and uncommon manner is awesome. wish you humans create greater such examine publications in close to future for our convenience.
Where can I find C2020-002 Actual Questions questions?
I wanted to start my own IT business but before it, C2020-002 course was necessary for my business, so I decide to get this certificate. When I took the admission for C2020-002 certification and took lectures I didnt understand anything. After some query I reached at killexams.com website and learnt from their and when my C2020-002 exam came I did well as compare to those students who took lectures and prepared from C2020-002 study guide from this website. I recommend this website to all. I also thank to the employees of this website.
I got extraordinary Questions bank for my C2020-002 examination.
A portion of the classes are extraordinarily intricate but I understand them utilizing the killexams.com Q&A and examination Simulator and solved all questions. basically attributable to it; I breezed via the test horribly essentially. Your C2020-002 dumps Product are unmatchable in exceptional and correctness. all of the inquiries to your object were in the checkas well. i was flabbergasted to check the exactness of your fabric. a lot obliged another time for your help and all theassist that you provided to me.
attempt out those actual C2020-002 modern-day and up to date dumps.
Have exceeded C2020-002 examination with killexams.com questions solutions. killexams.com is a hundred% reliable, most of the questions had been similar to what I were given on the exam. I neglected some questions just because I went blankand didnt consider the solution given within the set, but in view that I got the rest proper, I passed with top rankings. So my recommendation is to research everything you get on your training p.c. from killexams.com, this is all you want to pass C2020-002.
Dont waste your time on searching internet, just cross for those C2020-002 Questions and solutions.
The killexams.com material is simple to understand and enough to prepare for the C2020-002 exam. No other study material I used along with the Dumps. My heartfelt thanks to you for creating such an enormously powerful, simple material for the tough exam. I never thought I could pass this exam easily without any attempts. You people made it happen. I answered 76 questions most correctly in the real exam. Thanks for providing me an innovative product.
You just need a weekend to prepare C2020-002 exam with these dumps.
I wound up the exam with a satisfying 84% marks in stipulated time. thank you very plenty killexams. via and by, it become tough to do top to bottom look at intending with a full-time paintings. At that factor, I became to the Q&A of killexams. Its concise answers helped me to see a few complex topics. I selected to take a seat for the examination C2020-002 to reap in addition advancement in my career.
No waste of time on searhching internet! determined precise source of C2020-002 Q&A.
I cleared C2020-002 exam with high marks. Every time I had registered with killexams.com which helped me to score more marks. Its great to have help of killexams.com question bank for such type of exams. Thanks to all.
Do you need Latest Braindumps of C2020-002 exam to pass the exam?
C2020-002 QAs have stored my life. I didnt sense assured in this region and Im satisfied a chum has informed about killexams.com IBM package with me some days before the exam. I want i'd buy earlier, it'd have made things much easier. i thought that I passed this C2020-002 exam very early.
Great source of great real questions, accurate answers.
This is fantastic, I passed my C2020-002 exam last week, and one exam earlier this month! As many people point out here, these brain dumps are a great way to learn, either for the exam, or just for your knowledge! On my exams, I had lots of questions, good thing I knew all the answers!!
save your time and money, examine these C2020-002 Q&A and take the exam.
i've these days passed the C2020-002 exam with this bundle. that is a great answer if you need a quick yet dependable coaching for C2020-002 examination. this is a expert level, so expect that you nevertheless need to spend time gambling with Q&A - practical enjoy is fundamental. yet, as far and exam simulations cross, killexams.com is the winner. Their checking out engine clearly simulates the examination, such as the unique question types. It does make things less complicated, and in my case, I trust it contributed to me getting a one hundred% score! I could not consider my eyes! I knew I did nicely, but this became a marvel!!
C2020-002 Certification Brain Dumps Source : IBM Algo Financial Modeler Developer Fundamentals
Test Code : C2020-002
Test Name : IBM Algo Financial Modeler Developer Fundamentals
Vendor Name : IBM
Q&A : 60 Real Test Questions/Answers
Killexams NYSTCE exam dumps | Killexams Adwords-Search practice questions | Killexams 77-420 practical test | Killexams ST0-118 real questions | Killexams 000-052 brain dump | Killexams 70-413 exam questions | Killexams 310-019 real test | Killexams COG-142 sample questions | Killexams HP2-H32 braindump | Killexams LOT-988 test prep | Killexams HP0-045 past exams | Killexams 925-201b Q&A | Killexams 190-956 Practice Test | Killexams 300-360 practice questions | Killexams 190-832 sample test | Killexams 9A0-313 test questions | Killexams BCP-211 | Killexams 190-610 | Killexams 630-005 | Killexams MB2-704 |
3COM [8 Certification Exam(s) ]
AccessData [1 Certification Exam(s) ]
ACFE [1 Certification Exam(s) ]
ACI [3 Certification Exam(s) ]
Acme-Packet [1 Certification Exam(s) ]
ACSM [4 Certification Exam(s) ]
ACT [1 Certification Exam(s) ]
Administrat [1 Certification Exam(s) ]
Admission-Tests [12 Certification Exam(s) ]
ADOBE [90 Certification Exam(s) ]
AFP [1 Certification Exam(s) ]
AICPA [1 Certification Exam(s) ]
AIIM [1 Certification Exam(s) ]
Alcatel-Lucent [13 Certification Exam(s) ]
Alfresco [1 Certification Exam(s) ]
Altiris [3 Certification Exam(s) ]
American-College [2 Certification Exam(s) ]
Android [4 Certification Exam(s) ]
APC [2 Certification Exam(s) ]
APICS [1 Certification Exam(s) ]
Apple [69 Certification Exam(s) ]
Arizona-Education [1 Certification Exam(s) ]
ARM [1 Certification Exam(s) ]
Aruba [6 Certification Exam(s) ]
ASIS [2 Certification Exam(s) ]
ASQ [3 Certification Exam(s) ]
ASTQB [6 Certification Exam(s) ]
Autodesk [2 Certification Exam(s) ]
Avaya [85 Certification Exam(s) ]
Axis [1 Certification Exam(s) ]
Banking [1 Certification Exam(s) ]
BEA [5 Certification Exam(s) ]
BICSI [2 Certification Exam(s) ]
BlackBerry [17 Certification Exam(s) ]
BlueCoat [2 Certification Exam(s) ]
Business-Objects [11 Certification Exam(s) ]
Business-Tests [4 Certification Exam(s) ]
CA-Technologies [20 Certification Exam(s) ]
Certification-Board [9 Certification Exam(s) ]
Certiport [3 Certification Exam(s) ]
CheckPoint [31 Certification Exam(s) ]
CIPS [4 Certification Exam(s) ]
Cisco [270 Certification Exam(s) ]
Citrix [35 Certification Exam(s) ]
CIW [17 Certification Exam(s) ]
Cloudera [10 Certification Exam(s) ]
Cognos [19 Certification Exam(s) ]
College-Board [2 Certification Exam(s) ]
CompTIA [33 Certification Exam(s) ]
ComputerAssociates [6 Certification Exam(s) ]
CPP-Institute [1 Certification Exam(s) ]
CWNP [12 Certification Exam(s) ]
Dassault [2 Certification Exam(s) ]
DELL [7 Certification Exam(s) ]
DMI [1 Certification Exam(s) ]
ECCouncil [18 Certification Exam(s) ]
ECDL [1 Certification Exam(s) ]
EMC [122 Certification Exam(s) ]
Enterasys [13 Certification Exam(s) ]
Ericsson [5 Certification Exam(s) ]
Esri [2 Certification Exam(s) ]
ExamExpress [15 Certification Exam(s) ]
Exin [39 Certification Exam(s) ]
ExtremeNetworks [3 Certification Exam(s) ]
F5-Networks [19 Certification Exam(s) ]
Filemaker [9 Certification Exam(s) ]
Financial [35 Certification Exam(s) ]
Fortinet [10 Certification Exam(s) ]
Foundry [6 Certification Exam(s) ]
Fujitsu [2 Certification Exam(s) ]
GAQM [7 Certification Exam(s) ]
Genesys [4 Certification Exam(s) ]
Google [4 Certification Exam(s) ]
GuidanceSoftware [2 Certification Exam(s) ]
H3C [1 Certification Exam(s) ]
HDI [9 Certification Exam(s) ]
Healthcare [3 Certification Exam(s) ]
HIPAA [2 Certification Exam(s) ]
Hitachi [27 Certification Exam(s) ]
Hortonworks [1 Certification Exam(s) ]
Hospitality [2 Certification Exam(s) ]
HP [712 Certification Exam(s) ]
HR [1 Certification Exam(s) ]
HRCI [1 Certification Exam(s) ]
Huawei [20 Certification Exam(s) ]
Hyperion [10 Certification Exam(s) ]
IBM [1491 Certification Exam(s) ]
IBQH [1 Certification Exam(s) ]
ICDL [6 Certification Exam(s) ]
IEEE [1 Certification Exam(s) ]
IELTS [1 Certification Exam(s) ]
IFPUG [1 Certification Exam(s) ]
IIBA [2 Certification Exam(s) ]
IISFA [1 Certification Exam(s) ]
Informatica [2 Certification Exam(s) ]
Intel [2 Certification Exam(s) ]
IQN [1 Certification Exam(s) ]
IRS [1 Certification Exam(s) ]
ISACA [4 Certification Exam(s) ]
ISC2 [6 Certification Exam(s) ]
ISEB [24 Certification Exam(s) ]
Isilon [4 Certification Exam(s) ]
ISM [6 Certification Exam(s) ]
iSQI [7 Certification Exam(s) ]
Juniper [54 Certification Exam(s) ]
Legato [5 Certification Exam(s) ]
Liferay [1 Certification Exam(s) ]
Lotus [66 Certification Exam(s) ]
LPI [21 Certification Exam(s) ]
LSI [3 Certification Exam(s) ]
Magento [3 Certification Exam(s) ]
Maintenance [2 Certification Exam(s) ]
McAfee [8 Certification Exam(s) ]
McData [3 Certification Exam(s) ]
Medical [25 Certification Exam(s) ]
Microsoft [228 Certification Exam(s) ]
Mile2 [2 Certification Exam(s) ]
Military [1 Certification Exam(s) ]
Motorola [7 Certification Exam(s) ]
mySQL [4 Certification Exam(s) ]
Network-General [12 Certification Exam(s) ]
NetworkAppliance [35 Certification Exam(s) ]
NI [1 Certification Exam(s) ]
Nokia [2 Certification Exam(s) ]
Nortel [130 Certification Exam(s) ]
Novell [37 Certification Exam(s) ]
OMG [9 Certification Exam(s) ]
Oracle [232 Certification Exam(s) ]
P&C [1 Certification Exam(s) ]
Palo-Alto [3 Certification Exam(s) ]
PARCC [1 Certification Exam(s) ]
PayPal [1 Certification Exam(s) ]
Pegasystems [10 Certification Exam(s) ]
PEOPLECERT [4 Certification Exam(s) ]
PMI [15 Certification Exam(s) ]
Polycom [2 Certification Exam(s) ]
PostgreSQL-CE [1 Certification Exam(s) ]
Prince2 [6 Certification Exam(s) ]
PRMIA [1 Certification Exam(s) ]
PTCB [2 Certification Exam(s) ]
QAI [1 Certification Exam(s) ]
QlikView [1 Certification Exam(s) ]
Quality-Assurance [7 Certification Exam(s) ]
RACC [1 Certification Exam(s) ]
Real-Estate [1 Certification Exam(s) ]
RedHat [8 Certification Exam(s) ]
RES [5 Certification Exam(s) ]
Riverbed [8 Certification Exam(s) ]
RSA [13 Certification Exam(s) ]
Sair [8 Certification Exam(s) ]
Salesforce [3 Certification Exam(s) ]
SANS [1 Certification Exam(s) ]
SAP [78 Certification Exam(s) ]
SASInstitute [15 Certification Exam(s) ]
SAT [1 Certification Exam(s) ]
SCO [9 Certification Exam(s) ]
SCP [6 Certification Exam(s) ]
SDI [3 Certification Exam(s) ]
See-Beyond [1 Certification Exam(s) ]
Siemens [1 Certification Exam(s) ]
Snia [6 Certification Exam(s) ]
SOA [15 Certification Exam(s) ]
Social-Work-Board [1 Certification Exam(s) ]
SUN [63 Certification Exam(s) ]
SUSE [1 Certification Exam(s) ]
Sybase [17 Certification Exam(s) ]
Symantec [132 Certification Exam(s) ]
Teacher-Certification [3 Certification Exam(s) ]
The-Open-Group [8 Certification Exam(s) ]
TIA [3 Certification Exam(s) ]
Tibco [18 Certification Exam(s) ]
Trend [1 Certification Exam(s) ]
TruSecure [1 Certification Exam(s) ]
USMLE [1 Certification Exam(s) ]
VCE [5 Certification Exam(s) ]
Veeam [2 Certification Exam(s) ]
Veritas [25 Certification Exam(s) ]
Vmware [51 Certification Exam(s) ]
Wonderlic [1 Certification Exam(s) ]
XML-Master [3 Certification Exam(s) ]
Zend [5 Certification Exam(s) ]