Thwart Fraud Using Graph-Enhanced Machine Learning and AI
5 likes1,205 views
This webinar will discuss using graph-enhanced machine learning and AI to thwart fraud. On February 6th, Scott Heath from Expero and Amy Hodler from Neo4j will discuss how graph databases can be used to identify patterns and relationships in complex transactional data to detect fraud. The webinar is part of a series that will also cover building intelligent fraud prevention systems using machine learning and graphs, and obtaining funding for graph-enhanced fraud solutions.
![Graph + ML Fraud Analysis System
X[n]
K
N-1
Extract financial history data
AI - Analysis says: this
company is committing
fraudulent transactions
Clustering - Find corporate look-alikes
for fraud analysis
1
2
3
4
5](https://image.slidesharecdn.com/neo4jbusiness-fraudv14final-180207230305/85/Thwart-Fraud-Using-Graph-Enhanced-Machine-Learning-and-AI-38-320.jpg)
Thwart Fraud Using Graph-Enhanced Machine Learning and AI
- 1. WEBINAR: Thwart Fraud Using Graph-Enhanced Machine Learning and AI February 6th Scott Heath, Expero Amy Hodler, Neo4j © 2017 Expero, Inc. and Neo4j,Inc. All Rights Reserved
- 2. Who We Are SCOTT HEATH Graph Practice Manager, Expero Scott.Heath@experoinc.com @experoinc AMY HODLER Analytics Program Manager, Neo4j Amy.Hodler@neo4j.com @amyhodler
- 3. 500+ 7/10 12/25 8/10 53K+ 100+ 250+ 450+ Adoption Top Retail Firms Top Financial Firms Top Software Vendors Customers Partners • Creator of Neo4j Graph Platform • ~200 employees • HQ in Silicon Valley, other offices include London, Munich, Paris and Malmö • $80M in funding from Fidelity, Sunstone, Conor, Creandum, and Greenbridge Capital • Over 10M+ downloads, • 250+ enterprise subscription customers with over 50% with $1B and more in revenue Neo4j - The Graph Company Ecosystem Startups in program Enterprise customers Partners Meetup members Events per year Industry’s Largest Dedicated Investment in Graphs
- 4. © 2017 Expero, Inc. All Rights Reserved 4Behind every great product is a great team. Let’s build something great together. Expero ‘Certified’ NEO4J Professional Services ● Custom Application Experts : Full Lifecycle Applications ● Datastax Production Toolkits - Save Time and money ✓ Product Innovation ✓ Software Modernization ✓ Machine Learning & AI ✓ Graph Applications ✓ Neo4J ✓ Spark ✓ Solr
- 5. NEO4J + EXPERO = COMPLETE ENTERPRISE SYSTEMS Set of methods, tools & protocols to build software applications U and Visualization enabling users to perform self-service Application Layers - Micro services - REST Server End User Open Source, COTS & Custom - React, Angular - Keylines, Linkurious - D3 Full Applications: • Custom Industry function • Dashboard • Reporting • Visualize Data Structured/Unstructured data Extract Source Data Full Enterprise & Standardized Data Extract & transform source data to meet mission needs, load data into unified database Open Source & COTS Resolve & persist data; include multiple software & hardware elements Legacy + Custom + Industry Data and Platforms Source Data - Legacy - RDBMS - Analytics - Data Lakes - Data Marts SOURCE DATA EXTRACT, TRANSFORM & LOAD (ETL) DATA & MIXED DATA MODEL GRAPH DATA & PLATFORM An entity-centric, schema less, and self describing information management system APPLICATION LAYERS USER INTERFACE (UI) APIs PRESENTATION LOGIC DATA Source Apps - SFDC - SAP - Oracle
- 6. 6 Join Us - Webinar Series (Save the Dates !) Thwart Fraud Using Graph-Enhanced ML & AI You Are Here Build Intelligent Fraud Prevention with ML and Graphs Overview Technical Aspects Understand Business Impact Feb 13 9:00 PST / 12:00 EST Lock Down Funding for Graph-Enhanced Fraud Solutions Get Funding Feb 20 9:00 PST / 12:00 EST
- 7. What We Do
- 8. Neo4j — Changing the World ICIJ used Neo4j to uncovered the world’s largest journalistic leak up date, The Panama Papers, exposing criminals, corruption and extensive tax evasion. The US space agency uses Neo4j for their “Lessons Learned” database to connect information to improve searchability effectiveness in space mission. eBay uses Neo4j to enable machine learning through knowledge graphs powering “conversational commerce” Product RecommendationsFraud Detection Knowledge Graphs
- 9. 9 Harnessing Connections Drives Business Value Enhanced Decision Making Hyper Personalization Massive Data Integration Data Driven Discovery & Innovation Product Recommendations Personalized Health Care Media and Advertising Fraud Prevention Network Analysis Law Enforcement Drug Discovery Intelligence and Crime Detection Product & Process Innovation 360 view of customer Compliance Optimize Operations Connected Data at the Center AI & Machine Learning Price optimization Product Recommendations Resource allocation Digital Transformation Megatrends
- 10. Reverberations of Fraud |
- 11. • Increasing Unseen Costs • Organized & Adaptive Increasing Sophistication of Fraud Identity fraudsters bilked ~$28 billion from 30 million U.S. consumers in 2017* *Source: Nilson Report - Oct.com
- 12. • Increasing Unseen Costs • Organized & Adaptive • Societal Impact Increasing Sophistication of Fraud $100B+ Estimated Illegal Opioid Insurance Fraud
- 13. • Money Laundering • Credit Card • Check • Identity Theft • Combinations • With nuances in each industry • Insurance, retail, telecom... Many Faces of Fraud But there are commonalities • ‘Smurfing’ • Transactions • Actors • Locations • Devices Which means there a common traits, data, and patterns (or anti-patterns!) that can be analyzed!
- 14. The Graph Advantage • Pattern matching • Relationship & association analysis • Real-time monitoring and decisions • Reflexive to dynamic changes
- 16. Think Different ….. Think Graph
- 17. Forecast Complex Behavior and Prescribe Action Extract Structure and Model Processes
- 18. “There is No Network in Nature that we know of that would be described by the Random network model.” - Albert-László Barabási
- 19. Small-World High local clustering and short average path lengths. Hub and spoke architecture. Scale-Free Hub and spoke architecture preserved at multiple scales. High power law distribution. Random Average distributions. No structure or hierarchical patterns.
- 20. Averages Approach on Structured Data?NodeswithkLinks Number of Links (k) Average Distribution - Random - Most nodes have the same number of links No highly connected nodes NodeswithkLinks Number of links (k) Power Law Distribution - Scale-Free - Many nodes with only a few links A few hubs with a large number of links Source: Network Science - Barabasi
- 21. NodeswithkLinks Number of Links (k) Average Distribution - Random - Art: Ulysses and the Sirens – Herbert James Draper Most nodes have the same number of links No highly connected nodes You’ll Also Miss the Structure Hidden in Your Networks - Scale-Free - - Small World - Averages Approach on Structured Data?
- 22. Hierarchies On Stage Business Processes Behind the Scene Data Structure Linear Supply Chain / Decisions Information
- 23. On Stage Behind the Scene Organizations Multi-related Processes Knowledge Business Processes Data Structure
- 24. Structures Can Hide Source: “Communities, modules and large-scale structure in networks“ - Mark Newman Source: “Hierarchical structure and the prediction of missing links in networks”; ”Structure and inference in annotated networks” - A. Clauset, C. Moore, and M.E.J. Newman.
- 25. Reality
- 26. Older Methods: Slow, Costly and Painful
- 27. A Better Way …. Graph is Good
- 28. LOGICAL FLOW: SQL → Customer Applications SOURCE DATA DATA MAPPING - SOURCE MAP ENTITY RESOLUTION (ER) DATA, SEARCH & ANALYTICS PLATFORM APPLICATION PROGRAMMING INTERFACES (API) USER INTERFACE (UI) PRESENTATION LOGIC DATA PERSON Name / DOB / Products PERSON Name / DOB / Address COMPANY Shipper / Phone Number SOCIAL NETWORK SHIPPING ENTITIES FRAUD CUSTOMER 360 SUPPLY CHAIN RECOMMENDATIONS Map to Source Microservices - API Open Source & Custom Resolve and persist entities within and across datasets Use ML or Custom Algorithms An entity-centric, schemaless view, and self describing information management system Extract & transform or Create ‘Map’ of data - Federated Data Mapped Set of methods, tools, and protocols to build software applications Visualization tool enabling mission users to perform self-service data analysis Structured and unstructured data (e.g. social media, raid data) SQL, Triple Store, Hadoop, etc COMPANY Shipper / Address >>>>>>>>>> >>>>>>>>>> >>>>>>>>>> Source Data Schema Denormalized & Standardized Data >>>>>>>>>> COMPANY PERSON APIs PERSON Master Data Management Machine Learning Machine Learning
- 29. Graph databases store data based on relationships, rather than transactions Used For: Data analytics systems connecting disparate structured or unstructured data Graph Database Used For: Transactional systems with structured data Traditional Database Person FriendPerson_Friend Graphs are suited for environments where the connections between data points are just as important as the data points themselves.
- 30. ONBOARD name: “Dan” born: May 29, 1970 twitter: “@dan” name: “Ann” born: Dec 5, 1975 since: Jan 10, 2011 Name: “Consay” Headquarters: “NJ” Nodes • Can have Labels to classify • Labels have native indexes MARRIED TO LIVES WITH TRADES PERSON PERSON 30 Property Graph: Nodes + Relationships Company Relationships • Relate nodes by type and direction • Can have weights Properties • Attributes of Nodes & Relationships • Stored as Name/Value pairs • Can have indexes and composite indexes amount: $9,500
- 31. Exploring The Data Graph Gives us the Horsepower to See Differently Dependencies • Failure chains • Order of operation Matching / Categorizing Highlight variant of dependencies Clustering Finding things closely related to each other (friends, fraud) Flow / Cost Find distribution problems, efficiencies Similarity Similar paths or patterns Centrality, Search Which nodes are the most connected or relevant
- 32. Visualization and Interaction “Forests-of-Forests View” - Overwhelming Data “Tree-Leaf View” - Very Close In
- 33. A Better Way ... Graph + ML is Great
- 34. Intersection of Graphs & ML & Visualization
- 35. Graph is the Power, ML is the Force Multiplier ● Drive Action Fast ○ Recommend ‘Suspected’ items and flag them for investigation ○ Based on 10 Years of history Suggest new areas for investigation - i.e. ACH patterns show where to look ● Avoid Revenue Loss ○ Predict potential patterns and potential areas for investigation ● New Insights - ‘Treasure Map’ ○ Customer Clustering and Similarity ○ Campaigns: React to found data ○ Intelligent graph insight
- 36. Graph Enhanced ML & AI Knowledge Graphs Provide Rich Context for AI AI Visibility Human-Friendly Graph Visualization Graph Accelerated ML & AI Development Quickly Evaluate Datasets and Features for Extraction Graph Execution of AI Operationalize Real-Time OLAP and Monitoring Graph Enriched Data Preprocess and Augment Machine Learning Data Connected Source of Truth Data Lineage for ML System of Record for AI Decisions
- 37. Interactive Visualization ● Expert + Interactive ● Visualize potential risks from any other source: ○ pattern monitoring ○ machine learning ○ other LOB ● Rich Visualizations ○ identify “emerging risk connections” ○ connect-the-dots across risk cases AI + Machine Learning ● Semi -> Fully Automated ● Looking for risk patterns “we have not seen before” ● Looking for recurring patterns in transaction streams ● More effective at finding risks using lower number of data dimensions Cooperative/Hybrid Fraud Detection Stages Risk Pattern Monitoring ● Semi-> Fully Automated ● Looking for risk patterns “we have seen before” ● Code programs to look for specific patterns in transaction streams ● Can look at any number of dimensions in the data ● Fraud Rings constantly working to “crack the code” AnticipateGuard Discover
- 38. Graph + ML Fraud Analysis System X[n] K N-1 Extract financial history data AI - Analysis says: this company is committing fraudulent transactions Clustering - Find corporate look-alikes for fraud analysis 1 2 3 4 5
- 39. Identify ‘Hiding’ In plain sight and Attribution Fraud Potential
- 40. Fraud Detection - Transactional Fraud by Individuals Graph of individuals suspected fraud suspected fraud Machine learning highlights fraudulent transactions for bank review. Subgraph of transactions by an individual
- 41. Company Identity Lookalikes Low Risk High Risk Average Unstructured graph of companies Same graph, automatically clustered by their financial history similarities by an unsupervised learning algorithm.
- 42. A Better Way …. Graph + ML + Visualization is ...
- 43. Methods to Visualize - ML in Your Application? 1) Entity Link Analysis ○ Transactions : Amounts, Locations, Types of Goods, Types of stores, sizes of amount ○ Known Data : Matching against known previous fraud data 2) Graph Traversals ○ Entities or Actors : People, Companies, Goods and Services ○ Amounts : Odd amounts, small amounts with similar numbers ,repetitive locations ○ Known Data : Matching against data 3) Geospatial Viewing ○ Locations : Physical locations, corporate entities, ○ Devices: Mac Addresses, IP and device 4) Timeline Analysis ○ Reviewing all Events : Locations, Actors, Entities, Transactions, ○ Device Tracking: Mac Addresses, IP and device
- 44. Example: Use AI/ML + GRAPH To Create Action ML LINK: Tie Data to Action : Campaigns ● Activity ● Trends ● Loyalty ML PERSONALIZATION: Entity Link & Graph Traversal: ● Use History ● User Context ● Background
- 45. Example: AI + Graph Customer Journey ML Risk Analysis: ● Risk Factor ● Risk ● Sentiment AI Clustering: Entity Link & Graph Traversal: ● Activity ● Trends ● Background
- 46. Real-World Uses
- 47. DEMO - ART OF THE POSSIBLE HOW TO APPLY LIVE DEMOS
- 48. Rapid Prototype
- 49. Insight for Graph Methodology DISCOVERY INVENTION REALIZATION TRACK & MEASURE ONGOING SUPPORT PROOF OF CONCEPT PILOT TURN-KEY MVP DEVELOPMENT TECHNOLOGY LIFE CYCLE ASSESSMENTS : DIAGNOSE & PRESCRIBE - DATA, ARCHITECTURE, CODE, USER EXPERIENCE (Any Stage) SUPPORT - EXPERT SERVICES
- 50. Playbook: What are the Next Steps? Prototype Pilot Delivery Data Loading DSE Platform Data Discovery Craft Visualization Key Business Functions Build Rapid Pilot - Prototype Validate Business Case and Platform Technology ● Key Customer Functionality ● Graph Data Platform - Specifications ● Working Graph System ● Real Data Set Business Problem Go LiveDevelopmentDiscovery & Requirements Testing PLAY: Rapid Prototype
- 51. RAPID PILOT: See and Experience Your Data Web UI framework React Visualizations EXPERO GRAPH TOOLS + (Open Source) Graph Platform App Server (Generic Server) Provisioning EXPERO GRAPH TOOLS Ansible + Cloudburst Compute Cloud AWS EC2 Data Sources CUSTOMER Data or (Synthetic Data)
- 52. Join Us - Webinar Series Thwart Fraud Using Graph-Enhanced ML & AI You Are Here Build Intelligent Fraud Prevention with ML and Graphs Overview Technical Aspects Understand Business Impact Feb 13 9:00 PST / 12:00 EST Lock Down Funding for Graph-Enhanced Fraud Solutions Get Funding Feb 20 9:00 PST / 12:00 EST
- 53. Neo4j + Expero Complete Fraud Solutions PRESENTATION LOGIC DATA
- 54. Thank You! SCOTT HEATH Graph Practice Manager, Expero Scott.Heath@experoinc.com AMY HODLER Analytics Program Manager, Neo4j Amy.Hodler@neo4j.com @amyhodler www.Neo4j.com /use-cases/fraud-detection info@neo4j.com @neo4j www.ExperoInc.com /graph/graphs-are-everywhere info@experoinc.com @experoinc