De-Mystefying Predictive Analytics
2 likes1,559 views
The document outlines a talk on predictive analytics, discussing its differentiation from reporting tools, the predictive analytics process, and various real-world applications in retail and employee training. It emphasizes the importance of correlation over causation in predictive models, showcasing examples such as customer churn, employee training, and demand forecasting. Key components include problem identification, data measurement, model training, and the necessity for management commitment and creativity in implementation.
![K-Nearest Neighbours
Customer1 = [age=25, exper=1, income=49, family=4, CCAvg=1.6, education=UG,…]
Customer2 = [age=49, exper=19,income=34, family=3, CCAvg=1.5, education=UG,…]](https://image.slidesharecdn.com/demystefyingpredictiveanalyticsretechcon2012-120519103526-phpapp02/85/De-Mystefying-Predictive-Analytics-23-320.jpg)
![More predictive analytics methods:
based on distance
Customer1 = [age=25, exper=1, income=49, family=4, CCAvg=1.6, education=UG,…]
Customer2 = [age=49, exper=19,income=34, family=3, CCAvg=1.5, education=UG,…]](https://image.slidesharecdn.com/demystefyingpredictiveanalyticsretechcon2012-120519103526-phpapp02/85/De-Mystefying-Predictive-Analytics-25-320.jpg)
De-Mystefying Predictive Analytics
- 1. Galit Shmuéli SRITNE Chaired Prof. of Data Analytics De-mystifying Predictive Analytics
- 5. Today’s Talk 1. How predictive analytics differ from Reporting and other BI tools 2. The predictive analytics process 3. Examples of problems that can be tackled 4. Logic behind predictive analytics algorithms 5. Predictive Analytics for retail in India
- 7. Overall Behaviour Case Studies Past Present Future “Presonalized” Behaviour
- 8. Today’s Talk 1. How predictive analytics differ from Reporting and other BI tools 2. The predictive analytics process 3. Examples of problems that can be tackled 4. Logic behind predictive analytics algorithms 5. Predictive Analytics for retail in India
- 9. The Predictive Analytics Process Problem Identification Measurement Data Models Determine Draw sample, Data Mining Outcome and Split into algorithms Predictors training/holdout & Evaluation Deployment Re-evaluation More data
- 10. Today’s Talk 1. How predictive analytics differ from Reporting and other BI tools 2. The predictive analytics process 3. Examples of problems that can be tackled 4. Logic behind predictive analytics algorithms 5. Predictive Analytics for retail in India
- 11. Example 1: Personalized Offer Problem Who to Which What Identification target? coupon? medium? Measurement Data Models Outcome: redemption From similar past ? Predictors: customer, campaign Expected shop & product info (redeemers and gain per non-redeemers) offer sent Deployment (or not!) Re-evaluation More data
- 12. Example 2: Employee Training Problem Which employees to train? Identification Measurement Data Models Outcome: performance From past ? Predictors: employee & training efforts Expected training info (successes and gain per failures) employee Deployment (or not!) Re-evaluation More data
- 13. Example 3: Customer Churn Problem Identification Which members most likely not to renew? Membership renewal Measurement Data Models Outcome: renewal Past renewal ? Predictors: customer & campaigns Expected membership info (successes and gain per Deployment (or not!) failures) customer Re-evaluation More data
- 14. Example 4: Product-level demand forecasting Problem Weekly Identification forecasts per clothing item Measurement Outcome: month-ahead weekly forecasts of #units purchased per item Predictors: past demand for this & related items, special events, economic outlook, social media Deployment (or not!) Data Models Re-evaluation Historic info ? More data Expected gain
- 15. Example 5: COD Prediction Problem Predict payment Identification probability Measurement Data Models Outcome: pay/not Past deliveries ? Predictors: customer, (payments and Expected product, transaction info non-payments) gain per transaction Deployment (or not!) Re-evaluation More data
- 16. Today’s Talk 1. How predictive analytics differ from Reporting and other BI tools 2. The predictive analytics process 3. Examples of problems that can be tackled 4. Logic behind predictive analytics algorithms 5. Predictive Analytics for retail in India
- 17. Predictive Analytics: It’s all about correlation, not causation Every time they turn on the seatbelt sign it gets bumpy! Algorithms search for correlation between the outcome and predictors Different algorithms search for different types of structure
- 18. Example: Direct Marketing Maharaja Bank wants to run a campaign for current customers to purchase a loan They want to identify the customers most likely to accept the offer They use data from a previous campaign on 5000 customers, where 480 (9.6%) accepted
- 19. Data sample
- 20. Data Partitioning Training 4,000 customers Holdout 1,000 customers
- 21. Classification & Regression Trees No Yes No Yes No Yes
- 22. Regression Models Probability (Accept Offer) = function of b0 + b1 Age + b2 Experience + b3 Income + b4 CCAvg +… The Regression Model Input variables Coefficient Constant term -6.16805744 Age -0.0227915 Experience 0.03030424 Income 0.06047214 ZIP Code -0.00006691 Family 0.61913204 CCAvg 0.13191609 Mortgage 0.00016262 Securities Account -0.51986736 CD Account 4.10482931 Online -1.11415482 CreditCard -1.02319455 EducGrad 3.93598175 EducProf 4.01372194
- 23. K-Nearest Neighbours Customer1 = [age=25, exper=1, income=49, family=4, CCAvg=1.6, education=UG,…] Customer2 = [age=49, exper=19,income=34, family=3, CCAvg=1.5, education=UG,…]
- 24. Performance Evaluation: Holdout Data Predict each Overall Missed Targeted Error acceptors non- customer’s action acceptors Baseline: no offers 9.3% 9.3% 0.0% Holdout Tree 2.5% 12.9% 1.4% 1,000 customers Regression 4.3% 35.5% 1.1% K-NN 4.3% 41.9% 0.4% Different: Identify 20% of customers most likely to accept
- 25. More predictive analytics methods: based on distance Customer1 = [age=25, exper=1, income=49, family=4, CCAvg=1.6, education=UG,…] Customer2 = [age=49, exper=19,income=34, family=3, CCAvg=1.5, education=UG,…]
- 26. Where do the buzzwords fit in?
- 27. Big Data Cloud Computing Real-time data Unstructured Social Media data Mobile Data
- 28. Today’s Talk 1. How predictive analytics differ from Reporting and other BI tools 2. The predictive analytics process 3. Examples of problems that can be tackled 4. Logic behind predictive analytics algorithms 5. Predictive Analytics for retail in India
- 29. Step 1: Identify “classic” applications used by other companies
- 30. Step 2: Get Creative In India: Cash On Delivery Counter service Huge growth in ATMs Multiple languages Regional customer preferences Informative names Bargaining
- 31. What you’ll need Top management commitment Analytics team with close ties to all departments (IT, Marketing,…) understands the business and its goals creative and fearless is allowed to experiment (and fail) Data in a reachable place Software
- 32. Last Thought: Mindful Predictive Analytics “VIP syndrome” Predictive analytics for scaling-up to public white- glove treatment Predictive analytics for reducing the burden on consumers, employees etc. (less offers & overload)
- 33. Asia Analytics Lab @ ISB facebook.com/groups/asiaanalytics