Building the analytics Poly Analyst empowered enterprise Web
Building the analytics Poly. Analyst empowered enterprise Web Report Training Megaputer Intelligence www. megaputer. com © 2014 Megaputer Intelligence Inc.
Standard Outline. Functional Areas Company Marketing Customers Loss Prevention Customer Service Quality Assurance Risk Management Engineering Finance Human HR Resources Competitive Intelligence Competitors Suppliers Public Relations General Public Audit Legal Regulators
Typical Tasks - examples Outline • Social Media Data Analysis • Call Center Data Analysis • Survey Analysis • Incident Report Analysis • Fraud Detection • Subrogation/Litigation Prediction • Database Marketing • Sales Data Analysis
Social Media Data Analysis Outline Implemented in: § Marketing Dept Thousands of communications Key topics, sentiment & trends v Customer-driven insights Immediate & intelligent action v
Call Center Data Analysis Outline Implemented in: § Call Center Millions of phone call transcripts (text) Results of categorization offered in OLAP cubes v Objective, consistent results v Corporate awareness
Survey Analysis Outline Implemented in: § Vo. C § Vo. E Thousands of free text survey responses Easy to comprehend reports v Analysis of 100% of data v Responsive company
Incident Report Analysis Outline Implemented in: § Safety Department Thousands mixed incident reports Key issues and root causes v Improved Safety v Shortened response time
Fraud Detection Outline Implemented in: § SIU Millions of financial transactions Anomalies and potential fraud cases + Investigation v. Prevention/recovery v. Enhanced of losses system integrity
Subrogation Outline Prediction Implemented in: § Subrogation Dept Millions of free text claims notes Extracted key patterns + Subrogation potential v Predicting probable subros v Real time and retrospective
Database Outline. Marketing Implemented in: §Marketing Dept Millions of random prospects Increased response rate through better targeting v Increased response rate v Better allocation of resources
Cross-Sell Outline. Analysis Implemented in: § Call Center § Internet Store Real-time recommendations Millions of historical transactions v Significant increase in sales v Better customer experience
Benefits for the Enterprise Outline • Dramatic cost reduction • Increase in quality and speed of the analysis • Objective and uniform data-driven analysis • Discovery of even unexpected issues suggested by data • Automated monitoring of known problems • Timely discovery of newly developing issues • Utilization of 100% of available data: structured and text • Up-to-date reports for executives • Easy to use and maintain solution
Example: Warranty Fraud Detection • Dealer level • Identify and rank Dealers that have a large number of anomalies in submitted claims • Claim level • Determine and store for application efficient business rules automating payment decision on individual new claims
Finding Anomaly Thresholds and Suspect Dealers Machine Learning & Statistics New claim Claims Data Determine Nature of Distributions Business Rules for Claim Scoring Payment Decision Detect Thresholds & Anomalies Identify Systematic Anomalies Group Anomalies by Dealer Claim level List of Anomaly Ranked Dealers Dealer level
Analytics for Individual Claim Processing Submitted Claims Business Rules Discoveries Technology Assisted Audit Data & Text Analysis, Modeling and Scoring Pay Reject
Detecting Anomalies in Data
Finding Anomaly Thresholds Lower Threshold Upper Threshold
Distribution Analysis (Di. An) script
Tail Thresholds – for each Causal Part Tail Thresholds
Overall Anomalies in ESP claims
Dealer Total Loss by Anomaly
Dealers by Total Loss due to Anomalies
Share of Anomalies to Full Operation
Total Loss by Share of Anomalies
Dealers by Anomaly Type
Total Loss by # of Different Anomalies
Strongest Correlations: Dealers to Anomalies
Drilling down to Dealer 00601 Favorite Causal Parts
New Inputs for Predictive Modeling Add Extra Metrics through Text Mining • Feature Extraction • Entity & Event Extraction • Search for Patterns • Clustering • Categorization • Summary Creation • Sentiment Analysis
Key Steps of Text Mining Graphical Reports 1 Textual Data 2 3 4 Data Cleansing Data-driven Analysis Analyst-driven Analysis Enriching Structured Data OLAP cubes
Correct Typical Mechanic’s Abbreviations
Auto-Correct Misspells > 43, 000 misspells corrected!
Example: Car Repair Notes
Detect Near Duplicate Descriptions Near Duplicate Mechanic’s Notes
Dealer 00601 only replaces Wipers
Contacting Megaputer Questions?
- Slides: 36