An IBM Proof of Technology with IBM SPSS Modeler
Agenda Welcome and Introductions Introduction to Predictive Analytics IBM SPSS Modeler Overview Exercise: Predictive in 20 Minutes IBM SPSS Modeler Basics Break Predictive Analytics Methodology and Applications Exercise 1: Find Patterns and Groups Exercise 2: Understand the Past, Predict the Future Exercise 3: Optimize Decisions Integrating Predictive Analytics with Planning and BI Closing Discussion 2
Purpose of the workshop Introduction to predictive analytics Stimulate thinking about how predictive analytics would benefit your organization Demonstrate ease of use of powerful technology Get experience in doing predictive analytics Explore multiple predictive analytics techniques Understand role of predictive analytics in decision optimization 3
What is predictive analytics? Predictive Analytics helps connect data to effective action by drawing reliable conclusions about current conditions and future events Gareth Herschel, Research Director, Gartner Group Enabling businesses to use predictive models to exploit patterns found in historical data to identify potential risks and opportunities before they occur. 4
Top Five Benefits of Predictive Analytics How has your organization benefited from predictive analytics: Achieve competitive advantage New revenue opportunities Increased profitability Increased customer service Operational efficiencies Source: Ventana Research Predictive Analytics Benchmark Research 45% 44% 55% 52% 68% Related Research Points: Management (76%) has no doubts that predictive analytics has given its business the competitive edge. Small (81%) businesses are especially convinced that they have achieved competitive advantage by using predictive analytics. In the Public Sector 1 : To best position governments and citizens for desired results, four focus areas now require integrated execution and measurement: information, engagement model, digital platform and analytics competency. Leaders acknowledge both a perpetual information explosion and a persistent data paradox the management dilemma associated with having too much data and too little insight. 5 1 IBM Institute for Business Value, Opening up government: How to unleash the power of information for new economic growth
The Future Made Real: Predictive Analytics in Action 6
Areas for Predictive Analytics IBM Software Warranty Claims Product quality assurance Predictive maintenance to predict equipment failure Who are our best employees? How do we keep our best employees from leaving? Which prospects to recruit? Public Safety Crime analysis Predictive policing Money laundering Tax audits & collections Who are our best customers? Can we get more like that? What/why do they buy? Why do they leave? Network Intrusion Education Which students are likely to drop out and why? Who are the best students and how to recruit more of them? Which alumni are likely to make larger donations?... and many more 7
Predictive Analytics Brings Value Need: Understand which combinations of factors predict student retention in online programs Benefits: Predict with approximately 80% certainty whether a given student is going to drop out Helps APUS build programs and maintain conditions for maximum student retention Need: Predict which of its small business customers were at the highest risk of churn Benefits: 60% improvement in revenue retention rates Realizing millions of dollars in annualized revenue protection Need: Fraud losses accounted for 6-10% of premium costs for customers and operational inefficiencies from agents managing high- & low-risk claims Benefits: Improved claims processing by 70% Saved $2.4M in just 4 months Need: To drive revenue, enabled call center operatives to make relevant real-time offers based on incoming customer calls Benefits: Identify customers more efficiently Sales from the call center doubled within six months of launch 8
IBM SPSS Modeler at a Glance Comprehensive predictive analytics platform All types of users at multiple analytical abilities Integrated deployment, optimization and decision management capabilities A visual interface with built-in guidance Structured and unstructured data Deployed on a desktop or integrated within operational systems Bring predictive intelligence to a single user or an entire enterprise 9
Model Building for Users of Multiple Analytical Abilities One-click modeling for the business user Visual analytical workbench to expand and extend analysis 10
Exercise: Predictive in 20 Minutes Goal: Identify who has responded to a marketing campaign Approach: Why? Use a data extract from a CRM Define which fields to use Choose the modeling technique Automatically generate a model to identify who has responded Review results To save marketing cost and increase marketing response, identify those likely to respond and focus marketing efforts on those prospects. 11
Break - please return in 15 minutes 12
Modeling Techniques in IBM SPSS Modeler Technique Usage Algorithms Classification (or prediction) Segmentation Association Used to predict group membership (e.g., will this employee leave?) or a number (e.g., how many widgets will I sell?) Used to classify data points into groups that are internally homogenous and externally heterogeneous. Identify cases that are unusual Used to find events that occur together or in a sequence (e.g., market basket) Auto Classifiers, Decision Trees, Logistic, SVM, Time Series, etc. Auto Clustering, K- means, etc. Anomoly detection APRIORI, Carma, Sequence 13
Extend Capabilities through R R Integration R Build/Score, Process and Output node support Scale R execution by leveraging database vendor provided R engines Custom Dialog Builder for R Provides the ability to create new Modeler Algorithm nodes and dialogs that run R processes Makes R usable for nonprogrammers 14
Uncovering Patterns in Unstructured Data Text Analytics Natural Language Sentiment Analysis Entity Analytics Disambiguate identity People, places, things Social Network Analysis Uncover relationships Find leaders and followers 15
The importance of text Because people communicate with words, not numbers, it has become critical to be able to mine text for its meaning and to sort, analyse, and understand it in the same way that data has been tamed. In fact, the two basic types of information complement each other, with data supplying the what and text supplying the why. Source IDC: Text Analytics: Software s Missing Piece? 16
Text Analytics Uses natural language processing heuristic rules and statistical techniques to reveal conceptual meaning in text Extracts concepts from text and categorizes them Makes unstructured qualitative data more quantifiable, enabling the discovery of key insights from sources such as survey responses, documents, emails, call center notes, web pages, blogs, forums and more Brings repeatability to ongoing decision making 17 17
Entity Analytics Overview Identify Matching Entities Analysts Struggle to Match Entities, Especially From Diverse Sources Natural variability (Bob vs. Robert) Errors (transposed month and day) Fabrications (fake identities) Enhances Model Accuracy Model Against a Complete, Accurate Entity Multiple Applications Business: is this the same order the customer submitted? Fraud: is this the same person who already defaulted on a loan? Government: is this the same vehicle that was carrying illegal content last time? Policing: Is this the same person who called us before? 18
Social Network Analysis Processes CDR (Call Data Record) data companies to produce social analysis Focuses around identifying groups, leaders and probabilities that others will churn based on influence Enhances existing churn predictions of Modeler Expressed as two new nodes in the Sources Palette Group Analysis what are the groups in my data and who are the leaders Diffusion Analysis uses existing churn information to determine who else that churner is likely to influence to leave 19
Optimize your operational decision with decision management Optimized Decisions Business Rules + + Optimization Predictive Analytics Optimize Actions Within Resource Constraints, Aligning Execution With Strategy All Data Empower Real-time and Adaptive Decisions Accommodating Changing Conditions Provide Front-line Employees and Systems With Recommended Actions 20
Optimize Predictive Results & Business Rules vs Constraints Determine outcome using models and rules Optimize against constraints Model and compare scenarios 21
Methodology CRoss-Industry Standard Process Model for Data Mining Describes Components of Complete Data Mining Project Cycle Shows Iterative Nature of Data Mining Vendor and Industry Neutral Data mining is a key discipline for applying predictive analytics 22
IBM SPSS Modeler Workbench Basics 23
Exercises Find Patterns and Groups Understand the Past, Predict the Future Optimize Decisions Classify customers into groups based on underlying characteristics Model response to marketing offers using historical data, apply to new customers and deploy results Evaluate marketing campaign optimization scenarios built using predicted outcomes and business rules 24
Exercises Find Patterns and Groups Understand the Past, Predict the Future Optimize Decisions Classify customers into groups based on underlying characteristics Model response to marketing offers using historical data, apply to new customers and deploy results Evaluate marketing campaign optimization scenarios built using predicted outcomes and business rules 25
Find Patterns and Groups Goal: Create segments of customers Approach: Customer data from a database or CRM Define which fields to use Automatically generate a model to group customers Why? Better customer understanding (demographics, socio-economic etc) Tailored messages for each group/segment Personal and more relevant for consumers 26
Exercises Find Patterns and Groups Understand the Past, Predict the Future Optimize Decisions Classify customers into groups based on underlying characteristics Model response to marketing offers using historical data, apply to new customers and deploy results Evaluate marketing campaign optimization scenarios built using predicted outcomes and business rules 27
Understand the Past, Predict the Future Goal: Identify who is likely to respond to a marketing offer Approach: Why? Use a data extract from a CRM Extract concepts from the open ended comments in a customer survey Define which fields to use Choose the modeling technique Automatically generate a model to identify who is likely to respond Review results Apply to new customers Deploy results for use by marketing team Target those likely to respond to offers to increase revenue, cut costs Using unstructured data improves modeling accuracy and provides more insight 28
Exercises Find Patterns and Groups Understand the Past, Predict the Future Optimize Decisions Classify customers into groups based on underlying characteristics Model response to marketing offers using historical data, apply to new customers and deploy results Evaluate marketing campaign optimization scenarios built using predicted outcomes and business rules 29
Optimize Decisions Goal: Identify which optimization scenario to use Approach: Why? Optimize based on revenue expectation, cost and predicted response Modify parameters to create scenarios Run scenario simulations Review optimization output Compare scenarios on multiple dimensions Select best scenario Understand what will happen to revenue, costs, and other business outcomes based on optimization Confidently select best optimization scenario for deployment 30
Deployment Integrating with existing systems A call center agent submits customer information during an interaction The reaction to the offer is tracked and used to refine the model Based on the predictive model, a single offer is presented to the customer 31
Exercises Find Patterns and Groups Understand the Past, Predict the Future Optimize Decisions Classify customers into groups based on underlying characteristics Model response to marketing offers using historical data, apply to new customers and deploy results Evaluate marketing campaign optimization scenarios built using predicted outcomes and business rules 32
IBM SPSS Modeler Within the Business Analytics Ecosystem Predictive Customer Analytics Acquire Grow Retain Predictive Operational Analytics Manage Maintain Maximize Predictive Threat & Fraud Analytics Monitor Detect Control Data Collection Social Media Analytics Statistics Modeler Analytic Server & Catalyst Analytic Answers Collaboration and Deployment Services IBM Research Etc 33
Integration with IBM Cognos Business Intelligence 1) Leveraging BI, identify problem or situation needing attention Common Business Model Cognos BI 2) SPSS predictive analytics feed results back into the BI layer 3) Results widely distributed via BI for consumption by business Users 34
Integration with IBM Cognos TM1 TM1 Source and Export nodes Score and/or manipulate data and export directly to TM1 35
Integration with IBM Cognos TM1 Before After How does predictive analytics enhance planning? Using sophisticated forecasting algorithms including: ARIMA Exponential Smoothing Adding external predictors such as: Weather patterns Inflation data Unemployment data 36
Workshop takeaways Easy to use, visual interface Short timeframe to be productive with actionable results Does not require knowledge of programming language No proprietary data formats Open architecture Business results focused Leverages the investments already made in technology Cost effective solution that delivers powerful results across organization Full range of algorithms for your business problems Big Data enabled (Hadoop, SQL Pushback) End-to-end solution Data preparation through real time interactions Use structured, unstructured and semi-structured data Integrated portfolio for business analytics Scales from a single desktop to an enterprise deployments 37
Closing Discussion YOUR 38
39
Appendix Suggested Books 40
IBM SPSS Modeler Cookbook From Amazon.com IBM SPSS Modeler Cookbook by Keith McCormick, Dean Abbott, Meta S. Brown, Tom Khabaza, Scott R. Mutchler Paperback - 382 pages (October 2013) (also on Kindle) ISBN : 1849685460 Written by those who teach and have been working with IBM SPSS Modeler since its beginnings. Full of practical examples that span the full gamut of capabilities. 41
Data Mining Overview From Amazon.com Paperback: 512 pages Publisher: Wiley; 1 edition (December 28, 1999) Language: English ISBN-10: 0471331236 ISBN-13: 978-0471331230 ; Good introductory text on data mining for marketing from two top communicators in the field 42
Handbook of Statistical Analysis and Data Mining Applications Handbook of Statistical Analysis and Data Mining Applications Robert Nisbet, John Elder IV, and Gary Miner Academic Press (2009) ISBN-10: 0123747651 An excellent guide to many aspects of data mining including Text mining. 43
Data Mining Algorithms From Amazon.com Data Mining: Practical Machine Learning Tools and Techniques with Java Implementations by Eibe Frank, Ian H. Witten Paperback - 416 pages (October 13, 1999) Morgan Kaufmann Publishers; ISBN: 1558605525; Best book I ve found in between highly technical and introductory books. Good coverage of topics, especially trees and rules, but no neural networks. 44