Preventing Fraud with Through Analytics Satya Bhamidipati Data Scientist Business Analytics Product Group Copyright 2014 Oracle and/or its affiliates. All rights reserved. 2
Tax Fraud in Increasing 27% Increase in EITC *GAO Report $600B Impact between Federal and State tax *GAO Report $15.6B EITC Improper Payments *TIGTA Report 642,000 Incidents 38% increase since 2010 *TIGTA Report Copyright 2014 Oracle and/or its affiliates. All rights reserved. Oracle Confidential Internal/Restricted/Highly Restricted 3
Anatomy of a Refund Scam Top 5 Cyber Crimes Tax Refund Scam Copyright 2014 Oracle and/or its affiliates. All rights reserved. Oracle Confidential Internal/Restricted/Highly Restricted 4
Refund Process Process Tax returns and release refunds within weeks Once released funds cannot be traced. Quick turn around time. Copyright 2014 Oracle and/or its affiliates. All rights reserved. Oracle Confidential Internal/Restricted/Highly Restricted 5
Challenges to Identify and Stopping Fraud Volume of Fraud Resources Lack of Machine Learning Algorithms Criminals are creative Technology Copyright 2014 Oracle and/or its affiliates. All rights reserved. Oracle Confidential Internal/Restricted/Highly Restricted 6
ORACLE and TPS Founded by Joan Barr (IRS) Brian Bequette (Intuit) 2009 Developed Fraud Solutions Partnered with Oracle in 2012 Cloud and on PremSolution Copyright 2014 Oracle and/or its affiliates. All rights reserved. Oracle Confidential Internal/Restricted/Highly Restricted 7
TPS Solution Predictive Data Analytics Internal Data Validation Third-party Data Validation Fraud Modeling, Rules and Deny Listing (a.k.a. Bad Listing ) Year-over-year Return Analysis Return Attachment (PDF) Analysis Proven Proprietary Fraud Algorithms Identity Verification Copyright 2014 Oracle and/or its affiliates. All rights reserved. Oracle Confidential Internal/Restricted/Highly Restricted 8
TPS Fraud System Identifies refund fraud and ID theft before any refund is paid. Uses current and historical facts to predict future fraud Heat maps that show where fraud is occurring. Shows related filings Copyright 2014 Oracle and/or its affiliates. All rights reserved. Oracle Confidential Internal/Restricted/Highly Restricted 9
Sample TPS Findings 1. The fraud problem is getting worse. 2. Fraud is expanding to international sources. 3. Fraudsters are not only stealing identities, but are also creating them. 4. W2-to-return analysis checks are being circumvented 5. The per-return average fraudulent refund amount has increased. 6. Revenue impact due to fraudulent refunds paid is doubling each year. 7. Strong evidence of automated bot fraudulent filings Copyright 2014 Oracle and/or its affiliates. All rights reserved. Oracle Confidential Internal/Restricted/Highly Restricted 10
[ Our findings ] Fraud is getting worse Tax year 2011 High-volume, low dollar fraud 250 international IP hits 30,465hard fraudulent returns detected $7,336,548 in hard fraudulent refunds Tax year 2012 Lower volume, higher dollar fraud 4,413 international IP hits 9,161fraudulent hard returns detected $17,532,173 in hard fraudulent refunds Tax year 2013 High volume, high dollar fraud 9,096 international IP hits 32,983hard fraudulent returns detected $44,498,435 in hard fraudulent refunds Copyright 2014 Oracle and/or its affiliates. All rights reserved. 11
[ Fraud Interface ] Sample Results Copyright 2014 Oracle and/or its affiliates. All rights reserved. 12
[ Our findings ] What Does All this Mean Fraudsters getting smarter No single fraud detection mechanism is wholly effective. Best-practice methodologies should be utilized. Copyright 2014 Oracle and/or its affiliates. All rights reserved. 13
TPS - Oracle solution components Transmission to state Interactive dashboards Ad hoc analysis Mobile Scorecards And more Oracle Business Intelligence Enterprise Edition Analytics, reporting and interactive visualization TPS fraud detection logic Integrated prediction and cross source validations Oracle Endeca Information Discovery Investigative data exploration and analysis TPS fraud pattern discovery Validation against unstructured sources Oracle Database Enterprise Edition Advanced Analytics and Spatial & Graph Options TPS fraud prediction Models and scoring Structured and semi-structured data Agency and external data sources Attachments, documents, case notes, emails TEX T Unstructured Data Copyright 2014 Oracle and/or its affiliates. All rights reserved.
Oracle Advanced Analytics Data in the Database Oracle Database Added Algorithms User tables?x Visual Interface Database Compute Engine Copyright 2014 Oracle and/or its affiliates. All rights reserved. Oracle Confidential Internal/Restricted/Highly Restricted 15
Analytic Methods Classification Regression Anomaly Detection Attribute Importance Association Rules F1 F2 F3 F4 K Means Linear regression Naïve Bayes Neural Networks Support Vector Machines Singular Value Decomposition Principle Component Analysis Copyright 2014 Oracle and/or its affiliates. All rights reserved. Oracle Confidential Internal/Restricted/Highly Restricted 16
Oracle Data Miner GUI Easy to Use Oracle Data Miner GUI for data analysts Work flow paradigm Powerful Multiple algorithms & data transformations Runs 100% in-db Build, evaluate and apply models Automate and Deploy Generate SQL scripts for deployment Share analytical workflows Copyright 2014 Oracle and/or its affiliates. All rights reserved.
Oracle Strategy for R Provide high-performance, scalable R environment tightly integrated with Oracle RDBMS and Hadoop For R users For Database & Big Data developers Full access to Database and HDFS objects High performance and scalability for all R operations Scalable, Natively integrated machine learning algorithms Deploy R scripts and store R calculation results in Database or Hadoop Execute embedded R scripts containing any R algorithm or calculation Access stored R results in Database or Hadoop HDFS Retrieve R computation results in graphical formats like XML or PNG Integrate R results into BI Applications Copyright 2014 Oracle and/or its affiliates. All rights reserved.
Oracle Endeca Information Discovery Endeca Information Discovery Unified Querying Interactive Exploration App Composition Endeca Server Endeca Information Discovery (EID) helps organizations quickly explore all relevant data Combine structured & unstructured data from disparate systems Automatically organize information for search, discovery & analysis Rapidly assemble easy to use analysis applications Faceted Data Model Integration Enrichment Copyright 2014 Oracle and/or its affiliates. All rights reserved.
Interactive Exploration and Discovery Advanced Search Search look-ahead Spell-correction Data-driven filtering Faceted Navigation Select attributes, like a web site + + Visual Analysis Charting & crosstabs Geographic visualization Tag clouds Copyright 2014 Oracle and/or its affiliates. All rights reserved.
Analytics and OBIEE In-database data mining builds predictive models that predict customer behavior OBIEE s integrated spatial mapping shows where Customer most likely to be HIGHand VERY HIGH value customer in the future Copyright 2014 Oracle and/or its affiliates. All rights reserved.
Contacts Satya.Bhamidipati@oracle.com Brian.Bequette@taxprocessingsystems.com Sandy.Fitzpatrick@oracle.com Copyright 2014 Oracle and/or its affiliates. All rights reserved. Oracle Confidential Internal/Restricted/Highly Restricted 22