IBM SPSS Data Mining Workshop Agenda

© 2014 IBM Corporation
An IBM Proof of Technology
IBM SPSS Data Mining Workshop
Laila Fettah– Technical Sales Specialist Advanced Analytics
Robin van Tilburg – Business analytics Specialty Architect
30 oktober 2014

IBM Software
2
Welcome to the Technical Exploration Center
Introductions
Access restrictions
Restrooms
Emergency Exits
Smoking Policy
Breakfast/Lunch/Snacks – location and times
Special meal requirements?

IBM Software
3
Introductions
Please introduce yourself
Name and organization
Current integration technologies/tools in use
What do you want out of this Data Mining Workshop?

IBM Software
4
Agenda
10:00-10:10 Welcome and Introductions
10:10-11:00 Introduction to Predictive Analytics
11:00-11:30 Exercise: Navigating IBM SPSS Modeler
11:30-12:00 Exercise: Predictive in 20 Minutes
12:00-12:45 Lunch
12:45-13:30 Data Mining Methodology and Application
13:30-14:00 Exercise: Data Mining Techniques
14:00-14:30 Exercise: Deployment
14:30-14:45 Wrap-up

IBM Software
5
Objectives
Introduction to predictive analytics and data mining
Stimulate thinking about how data mining would benefit your organization
Demonstrate ease of use of powerful technology
Get experience in “doing” data mining
See examples of existing customers and their realized ROI/benefits

IBM Software
6
“I used to think my job was all about arrests. Chasing bad guys.”
“Now, we figure out where to send patrols to stop crime before it happens.”

IBM Software
7
Smarter Planet
The world is changing, enabling organizations to make faster, better-informed decisions
Digital technologies (sensors and other monitoring instruments) are being embedded into every object, system and process.
All the data generated by digital technology is providing intelligence to help us do things better, improving our responsiveness and our ability to predict and optimize for future events.
INTELLIGENT
INSTRUMENTED
INTERCONNECTED
In the globalized, networked world, people, systems, objects and processes are connected, and they are communicating with one another in entirely new ways.

IBM Software
8
With this change comes an explosion in information …
… Yet organizations are operating with blind spots
Inefficient Access
1 in 2 don’t have access to the information across their organization needed to do their jobs
Lack of Insight
1 in 3 managers frequently make critical decisions without the information they need
Inability to Predict
3 in 4 business leaders say more predictive information would drive better decisions
Variety of Information
Volume of Digital Data
Velocity of Decision Making
Source: IBM Institute for Business Value

IBM Software
9
Leverage Information To Drive Smarter Business Outcomes
Increase Revenue
Increase Productivity
Reduce Costs
Reduce Risk

IBM Software
10
“I used to think my job was all about arrests. Chasing bad guys.”
“Now, we figure out where to send patrols to stop crime before it happens.”

IBM Software
Door middel van data mining kan de politie de delen van hun jurisdictie rangschikken
11
Minst waarschijnlijk dat…
Meest waarschijnlijk dat…

IBM Software
12
Why is predictive analytics important to your organization?
“The median ROI for the projects that incorporated predictive technologies was 145%, compared with a median ROI of 89% for those projects that did not.”
–Source: IDC, “Predictive Analytics and ROI: Lessons from IDC’s Financial Impact Study”

IBM Software
13
SPSS Customers: Business Objectives
Attract the best customers
Retain profitable customers
Grow customer value
Manage Risk
Detect and prevent Non-Compliance
“What is the likelihood a prospect will respond?”
“What is the most likely next product for each customer?“
“Which customers are likely to leave?”
“What activities are likely to be fraudulent?”
“Which customers are likely to default on a loan?”

IBM Software
14
Enabling the Predictive Analytics Process
Connect & Capture
Analyse & Predict
Deliver & Act
Data Collection delivers an accurate view of customer attitudes and opinions
Predictive capabilities bring repeatability to ongoing decision making, and drive confidence in your results and decisions
Unique deployment technologies and methodologies maximize the impact of analytics in your operation

IBM Software
15
SPSS Predictive Analytics Software -- 4 Product Families
Data Collection (surveys) Delivers accurate view of customer attitudes & opinions
•IBM SPSS Data Collection Statistics Drives confidence in your results & decisions
•IBM SPSS Statistics
•IBM SPSS Text Analytics for Surveys (STAFS) Modeling (data mining) Brings repeatability to ongoing decision making
•IBM SPSS Modeler
•IBM SPSS Text Analytics (TA) Deployment (automation, scoring service, sharing, …) Maximizes the impact of analytics in your operation
•IBM SPSS Decision Management
•IBM SPSS Collaboration & Deployment Services

IBM Software
16
Predictive Modeling with Modeler

IBM Software
17
Predicting Customer Behavior
Marketing activities are driven by predicted customer behavior
Data Mining
Data on Historic and Present Customer Behavior
Predicted Customer Behavior
Enterprise Data Sources
Marketing Attitudinal Interaction Web Call-center Operational
Attrition risk
Potential value
Cross sell
B
Cross sell
A
Credit risk
Fraud risk

IBM Software
18
Definition of Data Mining
Finding patterns in your data that you can use to do your business better
Business-oriented discovery of patterns producing insight and a predictive capability which can be deployed widely
Process of autonomously retrieving useful information or knowledge (“actionable assets”) from large data stores or set
Predictive analysis helps connect data to effective action by drawing reliable conclusions about current conditions and future events.” Gareth Herschel, Research Director, Gartner Group

IBM Software
19
Statistical vs. Data Mining Approach
Top-Down Approaches:
Query, Search
Bottom-Up Approaches:
Data Mining, Text Mining
A Statistical Approach can involve a user forming a theory about a possible relationship in a database and converting that to a hypothesis and testing that hypothesis using a statistical method. It is a manual, user-driven, top- down approach to data analysis. Source DM Review
•The difference with data mining is that the interrogation of the data is done by the data mining method--rather than by the user. It is a data-driven, self- organizing, bottom-up approach to data analysis that works on large data sets.
* "Statistical Modeling: The Two Cultures," Leo Breiman, Statistical Science, 2001, Vol.16 (3), pp.199-231.

IBM Software
20
Data Mining: a Different Approach
Top-Down Query Search (OLAP, BI)
Bottom-Up Data Mining Text/Web Mining
Measurement (historical)
Prediction (future)
Business value
Facts Segments & Trends Predictions
Data
mining Which customer types are at risk and why? Which cities were they located in?
OLAP
How many subscribers did we lose?
Query &
Reporting
What should we offer this customer today?
Integrated
Analytical
Solutions

IBM Software
21
IBM SPSS Modeler
High performance data mining and text analytics workbench
Used for the proactive
•Identification of revenue opportunities
•Reduction of costs
•Increase in productivity
•Forecasting
Allows analytics to be repeated and integrated within business systems

IBM Software
22
IBM SPSS Modeler

IBM Software
23
IBM SPSS Modeler

IBM Software
24
Exercise: Predictive in 20 Minutes
Goal:
Identify who has cancelled their contract Approach:
Use a data extract from a CRM
Define which fields to use
Choose the modeling technique
Automatically generate a model to identify who has cancelled
Review results Why?
To prevent customers cancelling, by proactively identifying those likely to cancel before they do.

IBM Software
25

IBM Software
26

IBM Software
27
Data mining 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

IBM Software
28
Data Mining Considerations – CRISP-DM
28
Business Understanding
What is the goal, what are we trying to achieve?
Data Understanding/Preparation
Available data (structured/unstructured)
Relevant factors
Subject matter expertise
Modeling
Supervised vs. Unsupervised
Different types of models (NN vs. Rules)
Combining models (Meta modeling)
Deployment
Batch vs. Real-time
Production Automation
Scheduling
Champion – Challenger
Multi-step jobs, conditional logic
Governance
Version control
Security and auditing

IBM Software
29
Business Understanding
Business Problem
Telco Company has seen an increase in Customer Churn. Problems with the Current Process
Based on Analysis it is not clear what the factors drive churn. The business is in reactive mode vs. proactive. Business Need
The executives have asked the marketing department to identify the customers that are likely to churn and create an action plan to address the problem.

IBM Software
30
Data Understanding
Do we have historical data that describes our customer behavior?
–Yes, the data is available in the Enterprise Data Warehouse
Do we have historical data of the customers that have churned?
–Yes, we keep that historical data in the EDW as well.
What data do we need? Where is it located?
–Billing data, call data, payment data and demographics

IBM Software
31
Data Preparation
Aggregate the data so that we have one row for each account
Get the relevant attributes and calculate them if necessary Demographic data Call behavioral data Churn flag

IBM Software
32
Modeling
In this phase, various modeling techniques are selected and applied, and their parameters are calibrated to optimal
values. Typically, there are several techniques for the same data mining problem type. Some techniques have specific
requirements on the form of data. Therefore, going back to the data preparation phase is often necessary.

IBM Software
33
Evaluation

IBM Software
34
Questions Customer Ask That Modeler Helps Answer
Segment
–I know my customers aren’t all the same, but how?
Acquire
–What customer should I be going after?
–Where should I put my new store?
Grow
–I’ve got dozens of products to offer– how do I know the best mix to offer?
–I’m blanketing my customer base with offers, but my returns seem to be diminishing. What am I doing wrong?
Retain
–I wish I knew which customers were most likely to leave me for a competitor.
–I wish I knew which customers were the most profitable
Fraud/Risk
–I am spending a lot of time reviewing each claim, I wish there was a way of identifying which claims I should focus on.

IBM Software
35
“After a thorough investigation of the analytical solutions in the market, we selected IBM SPSS for its ease of use for the business users and the extensive insight it provides into customer behavior and profitability. The software generates results rapidly.”
— Paul Groenland
Project manager, database marketing Rabobank
Business challenge
Rabobank aims to strengthen its position as a market leader in financial services by further developing and expanding its relationship with its private and corporate customers.
Solution
Rabobank uses predictive analytics software from IBM SPSS to create and execute targeted direct marketing and lead generation campaigns. The quality of the leads is higher, so marketing campaigns are much more cost- efficient and effective
Benefits
Completion time for marketing campaigns has decreased, on average, by two to four weeks
The quality of the leads is higher, so marketing campaigns are much more cost-efficient and effective
Highly targeted support for local banks and advisors. By providing timely and targeted leads, they can quickly respond to changes and to individual customers’ wishes.
Rabobank

IBM Software
36
Zorg en Zekerheid Uses business analytics to target fraudulent insurance claims
The need:
Processing millions of healthcare records requires surgical precision. For this Netherlands health insurer, this level of efficiency was missing from the process of analyzing claims and invoices to catch fraudulent activity. Manually selecting the data on the basis of predefined risk indicators had proven to be both time- consuming and unreliable in catching those abusing the system.
The solution:
Zorg en Zekerheid deployed a predictive analytics software solution capable of analyzing larger quantities of data, discovering patterns automatically,and catching anomalies in the process with a sharper level of accuracy and efficiency. The software provides a simple, graphical interface to deliver robust data mining, advanced analytics and interactive visualization for business users.
What makes it smarter:
Propels the fraud investigation process to action within days, instead of multiple weeks, using predictive analytics. Enables lost money to be recovered.
Captures all relevant data, including hard-copy invoices, which the system scans and archives.
Aggregates millions of digitally submitted records from multiple data sources and media formats into a central database, so data can be cross-functionally structured and automatically analyzed.
“The analytics solution has doubled our financial results each year since 2007.” — Andor de Vries, Fraud Analyst, Zorg and Zekerheid
Solution component:
IBM® SPSS Modeler

IBM Software
37
Data Mining Methods
Unsupervised Learning – Input and outputs are unknown, finds useful patterns
Supervised Learning – Modeler specifies what to predict
Clustering
Associations / Sequences
Regression
•Exploratory data analysis
•Reveals natural groups within a data set
•Distance Measure: No prior knowledge about groups or characteristics
•Not always an end in itself
•Finds things that occur together
•Associations can exist between any of the attributes
•Discovers association rules in time-oriented data
•Find the sequence or order of the events
Customer Segmentation
Market Basket Analysis, Next logical purchase
Classification
•Predicts an fixed outcome based on a set of inputs.
•Modelers pre-defines input and outputs
Fraudulent insurance claim prediction

IBM Software
38
38
Unsupervised Learning - Cluster and Associate
Clustering
–An exploratory data analysis technique
–Reveals natural groups within a data set
–Distance Measure: No prior knowledge about groups or characteristics
–Not always an end in itself
Associations
–Finds things that occur together – ex: events in a crime incident
–Associations can exist between any of the attributes (no single outcome like Decision Trees)
Sequential Associations
–Discovers association rules in time-oriented data
–Find the sequence or order of the events

IBM Software
39
39
Supervised Learning - Classification
Neural Networks
–A technique for predicting outcomes based on inputs where the inputs are weighted on hidden layers
–Behaves similar to the neurons in your brain
–Powerful general function estimators
–Require minimal statistical or mathematical knowledge
Decision Trees and Rule Induction
–Classification systems that predict or classify
–Technique that shows the ‘reasoning’ – contrast with Neural Network
–Builds sets of easy to understand ‘If – Then’ Rules
–Eliminates factors that are unimportant Cat.%nBad52.01168Good47.99155Total(100.00)323Credit ranking (1=default) Cat.%nBad86.67143Good13.3322Total(51.08)165Paid Weekly/MonthlyP-value=0.0000, Chi-square=179.6665, df=1Weekly payCat.%nBad15.8225Good84.18133Total(48.92)158Monthly salaryCat.%nBad90.51143Good9.4915Total(48.92)158Age CategoricalP-value=0.0000, Chi-square=30.1113, df=1Young (< 25);Middle (25-35) Cat.%nBad0.000Good100.007Total(2.17)7Old ( > 35) Cat.%nBad48.9824Good51.0225Total(15.17)49Age CategoricalP-value=0.0000, Chi-square=58.7255, df=1Young (< 25) Cat.%nBad0.921Good99.08108Total(33.75)109Middle (25-35);Old ( > 35) Cat.%nBad0.000Good100.008Total(2.48)8Social ClassP-value=0.0016, Chi-square=12.0388, df=1Management;ClericalCat.%nBad58.5424Good41.4617Total(12.69)41Professional

IBM Software
40
Anomaly Detection
Anomalies
–Anomaly detection is an exploratory method
–Designed for quick detection of unusual cases or records that should be candidates for further analysis
–These should be regarded as suspected anomalies, which, on closer examination, may or may not turn out to be real
40

IBM Software
41
Disclaimer: Common Sense Check

IBM Software
42
Richmond Police Department Curbing crime with predictive analytics
The need: Facing a rising crime rate, the Richmond Police Department needed an efficient and cost-effective way to analyze crime data, assess public safety risks and make intelligent decisions about personnel deployment. The solution: The Department turned to IBM SPSS, to deploy a powerful predictive analytics tool that brings data from multiple sources into one data warehouse; discovers hidden relationships in the data; and automatically generates crime forecasts.
Analyzes extremely large datasets and predicts crime patterns, giving the Department intelligence it needs to curb crime
Enables the Department to be efficient about how, where and when to deploy patrol and tactical units
Demonstrates ability to reduce violent-crime rates (homicide rates dropped 32 % from 2006-2007 and an additional 40 % from 2007-2008)
“The big performance boost has been for my new guys on the streets. IBM SPSS essentially does the work that is gained only from experience.”
— Stephen Hollifield
Head of Technology Richmond Police Department
Solution components:
IBM SPSS Statistics
IBM SPSS Modeler
IBM Business Partner Information Builders
IBM Business Partner RTI International

IBM Software
43
Association
Classification
Segmentation
Exercises

IBM Software
44
Association
Classification
Segmentation
Exercises

IBM Software
45
Association model
Goal:
Identify what products are being sold together Approach:
Use a data extract from a transactional system
Visualize relationship between products
Generate association model
Identify next likely purchase
Create bundles to increase $ value

IBM Software
46
Association
Classification
Segmentation
Exercises

IBM Software
47
Segmentation model

IBM Software
48
Association
Classification
Segmentation
Hands on sessions

IBM Software
49
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?”

IBM Software
50
Text data and text analytics
Around 80% of data held within a company is in the form of unstructured text documents or records:
–Insurance claim notes
–Emails
–Call center logs,
–Reports
–Surveys
–Web pages
–Blogs
– …
Text Analytics connects unstructured text data to effective action by drawing reliable conclusions about current conditions and future events

IBM Software
51
IBM SPSS Text Analytics
Bring repeatability to ongoing decision making

IBM Software
52
Sentiment Analysis
Hundreds of customers reviews at a glance…

IBM Software
53
Text Mining
Free form notes entries
Linguistic Text Mining:
1.Language analysis
2.Concept extraction
3.Process types, frequencies, & patterns
Integrated structured and unstructured data ready for Predictive Text Analytics

IBM Software
54
Use Text Analytics results to Improve Predictive Models

IBM Software
55
RTL Nederland / InSites Consulting - Analyzing social media buzz to increase TV viewer involvement
The need:
RTL Nederland aimed to evaluate its television programs in the Dutch market and increase viewer satisfaction making use of online conversations. Therefore, RTL Nederland needed a way to analyze, interpret and successfully respond to audience feedback from social media sources.
The solution:
RTL Nederland worked with InSites Consulting to capture viewer opinions from user-generated comments on social media and other online buzz by using IBM predictive analytics software. This helps RTL Nederland to better understand audience needs and preferences, and hence increase viewer satisfaction and involvement. The obtained insight on viewer likes and dislikes allows RTL Nederland to optimize its product offering.
Analyzed the sentiment of over 71,000 online conversations about ‘X FACTOR’, providing RTL Nederland with a powerful tool to measure attitudes indirectly and quickly adapt the program accordingly
Captures unstructured data automatically from the web with sophisticated text analytics technology
Approaching the final episodes of the reality competition shows, online buzz on the program even increased by about 400 percent, which provided a very rich source of information about viewer opinions
“Collecting and analyzing feedback from social media is of great importance to RTL Nederland in order to offer programmes that are fully aligned with the target audience.”
— Emilie van den Berge, senior Research & Intelligence project leader, RTL Nederland

IBM Software
56
Classification model
Goal:
Identify who is likely to cancel their contract Approach:
Use a data extract from a CRM
Use open ended comments from call center
Extract concepts from the text
Choose the modeling technique
Automatically generate a model to identify who has cancelled
Identify customers at risk before they churn
Unstructured data can provide insight into customers actions and improve model accuracy

IBM Software
57
Association
Classification
Segmentation
Exercises

IBM Software
58
Deployment

IBM Software
59
Deployment
Goal:
Deploy a predictive model Approach:
Use the stream generated in the earlier session
Pass new data through the stream and ‘score’ the data
Identify those likely to cancel
Export an .xls file with 50 most likely to cancel Why?
Extend the reach of analytics in an organization
Allows analytics at the point of impact rather than being reactive

IBM Software
60
Based on the predictive model, a single offer is presented to the customer
A call center agent submits customer information during an interaction
The reaction to the offer is tracked and used to refine the model
Deployment – integrating with existing systems

IBM Software
61
Customer Example Customer Growth from Inbound Contacts
“I’m calling to get my information on my download limit”
Next Best Action : Recommend Broadband Unlimited
“Certainly, Mr. Watson. I’ll just get that for you right now… “
“Mr.Watson, you currently close to your 10GB monthly limit however, as a valued long-term customer, we’re able to make you an offer on unlimited broadband”

IBM Software
62
Deployment – integrating with Cognos BI
3) Results widely distributed via BI for consumption by business Users
Cognos BI
Common Business Model
1) Leveraging BI, identify problem or situation needing attention
2) SPSS predictive analytics feed results back into the BI layer

IBM Software
63
Modeler’s Unique Capabilities
Easy to Learn / Intuitive Visual Interface
–Visual approach - no programming
–Comprehensive range of data mining functions
–Flexible deployment options
Powerful Automated modeling
–Automated data preparation
–Multi model creation & evaluation
–Integrated analysis of text, web, & survey data
Open and scalable architecure
–Data mining within standard databases with SQL pushback support
–Maximized use of infrastructure with multithreading, clustering and use of embedded algorithms (in database mining)
–Integration with IBM technologies such as IBM Cognos Business Intelligence, Netezza and IBM InfoSphere Warehouse

IBM Software
64
Modeler Editions
IBM SPSS Modeler Professional
–Modeler Professional is a data mining workbench for the analysis of structured numerical data to model outcomes and make predictions that inform business decisions with predictive intelligence.
IBM SPSS Modeler Premium
–Modeler Premium allows organizations to tap into the predictive intelligence held in all forms of data. Modeler Premium goes beyond the analysis of structured numerical data alone and includes information from unstructured data such as web activity, blog content, customer feedback, e-mails, articles, and more to create the most accurate predictive models possible.

IBM Software
65
IBM SPSS Modeler Deployment Options
Client (Desktop)
–Access local files
–Connect to operational databases
–Connect to Cognos BI
–Processing performed on local installation
Client/Server
–Data operations/processing on server
–In-database data mining
–SQL pushback
–Modeler Batch
–SuSE Linux Enterprise Server 10 (zLinux)
–Inclusion in Smart Analytics System for Power (AIX)
65

IBM Software
66
Workshop Takeaways
Easy to use, visual interface
Short timeframe to be productive with actionable results
Does not require knowledge of programming language Business results focused
Cost effective solution that delivers powerful results across organization
Flexible licensing and deployment options
Full range of algorithms for your business problems End-to-end solution
Data preparation through real time interactions
Use structured, unstructured and survey data
Full suite of products, from data collection through deployment

IBM Software
67
Workshop Takeaways
Flexible architecture
Leverages the investments already made in technology
Does not require data in a proprietary format or DB
Structured and unstructured data
Open architecture (both inputs and outputs)
SQL Pushback

IBM Software
68 IBM SPSS Data Mining Workshop
Predictive analytics customer success
 “94% achieved a positive return on investment with an average
payback period of 10.7 months.”
 “Returns were achieved through reduced costs, increased productivity,
increased employee and customer satisfaction, and greater visibility.”
 “Flexibility, performance, and price were all key factors in purchase
decisions.”
Nucleas Research, An independent provider of Global Research and Advisory Services.
“30 “100% increase in Million Euro in new revenue”
campaign effectiveness”
“Reduced churn from 19 to 2%” “35% reduction in mailing cost,
2X response rate, 29% more
profit”

IBM Software
69
We appreciate your feedback. Please fill out the survey form in order to improve this educational event.
SIMPLIFIED CHINESE
HINDI
JAPANESE
ARABIC
RUSSIAN
TRADITIONAL CHINESE
TAMIL
THAI
FRENCH
GERMAN
ITALIAN
SPANISH
BRAZILIAN PORTUGUESE

IBM Software
70
IBM Business Solutions Center, La Gaude – october 2011
Thank You
Laila Fettah Client Technical Professional Advanced Analytics
IBM Johan Huizingalaan 765 1066 VH Amsterdam Tel: +31 (0)20 513 8950
Mobile: +31 (0)6 11 87 61 55 robin.van.tilburg@nl.ibm.com
Robin van Tilburg Client Technical Professional Advanced Analytics
IBM Johan Huizingalaan 765 1066 VH Amsterdam Tel: +31 (0)20 513 8371 Mobile: +31 (0)6 31 04 10 74 lailafettah@nl.ibm.com
Contact

IBM SPSS Data Mining Workshop Agenda

Recommended

Recommended

More Related Content

What's hot

What's hot (20)

Similar to IBM SPSS Data Mining Workshop Agenda

Similar to IBM SPSS Data Mining Workshop Agenda (20)

More from Daniel Westzaan

More from Daniel Westzaan (14)

Recently uploaded

Recently uploaded (20)

IBM SPSS Data Mining Workshop Agenda