Se ha denunciado esta presentación.
Utilizamos tu perfil de LinkedIn y tus datos de actividad para personalizar los anuncios y mostrarte publicidad más relevante. Puedes cambiar tus preferencias de publicidad en cualquier momento.

Behavioral Analytics for Financial Intelligence

1.412 visualizaciones

Publicado el

Predictive analytics and machine learning has led to new methods for modeling human behavior and cognition. These methods are collectively known as behavioral analytics, focusing on how and why individuals and groups take actions and respond to them. This presentation discusses how financial services organizations are learning to leverage behavioral analytics for a broad set of applications that span customer insight, fraud detection, compliance, and market/investment intelligence.

Publicado en: Datos y análisis

Behavioral Analytics for Financial Intelligence

  1. 1. John Liu PhD, CFA Behavioral Analytics for Financial Intelligence
  2. 2. 2 Machine Learning platform that understands human communication Nashville Washington New York London Goldman Sachs Credit Suisse In-Q-Tel Silver Lake National Security Financial Services Healthcare Data Science CONFIDENTIAL 2 DIGITAL REASONING | CONFIDENTIAL Prepared Exclusively for Intel Capital Digital Reasoning Trusted Cognitive Computing For A Better World
  3. 3. Financial Services FX Rate Manipulation • Collusion and manipulation • "Quid-pro- quo” • Clustered Bid/ Ask quotes • Order execution around the benchmark • Trading abuse • Deal related language • Abusive/dirty references • Improper placement and execution Wall Cross Violations • Insider trading • M&A leakage • Personal trade execution • Companies on Restricted list •  Email from/to sensitive countries •  Bribery & gift language • Presence of attachments • Deal related language & order execution • Public info to enhance KYC profiles • Email to/from economically sanctioned countries Conduct Risk Anti-Bribery Anti-Money Laundering Unauthorized Trading 3 Electronic Communications Surveillance
  4. 4. • Counter-terrorism & proliferation • Human geography, pattern of life • Suspicious Activity Reporting (SAR) • Criminal investigations • Regulatory oversight • Disaster response Public Sector 4 DoD & IC Law Enforcement Civilian
  5. 5. •  Abundance of data available that captures human behavior •  Understanding human behavior makes for better decision making •  Behavioral analytics transforming financial industry 5 3 Big Ideas
  6. 6. BIG data: recording the annals of human behavior (YouTube collects 500TB/day, mostly about cats) Data, Data, Data 6
  7. 7. Types of Data Structured Data = Any data that can be stored in a table Unstructured Data = Everything else! Categories of Unstructured Data: •  Text (emails, im chats, tweets, legal docs, etc…) •  Image (Facebook, Instagram, webpages) •  Video (Youtube, security cameras, drones) •  Audio (phone calls, voicemail, open mic) •  GPS/IoT streams •  Romance novels 7
  8. 8. What is Behavioral Analytics? 8 We are creatures of habit. Definition: Behavioral analytics quantify the effects of psychological, social, cognitive, and emotional factors on human decision-making
  9. 9. Behavioral Analytics Purpose 9 Events, Relationships and Anomalies (Friends, Supply Chain, Cyber) Eat   Burp   Sleep   Process Mining: Habits and Behavior (Shopping Habits, Buying Cycle) Preferences: Likes and Dislikes (Advertising, Regulatory Reporting)
  10. 10. Behavioral Analytics 10 Descriptive Did Identify & Track Behavior Anomaly Detection Fraud, Cyber, Insider Predictive Will Do Credit Scoring Risk Management CRM & Cross- Selling Diapers & Beer Prescriptive Want To Do Behavior Modification Gamification Fixing Cognitive Heuristics/Biases
  11. 11. Behavioral Analytics Aspects 11 Group/Entity •  Baseline assessment •  Cohort analysis Individuals •  Personal habits •  Cognitive heuristics & biases Process Discovery Event/ Funnel Analysis Path Analysis Cohort Analysis
  12. 12. 12 Behavioral Analytics for Financial Intelligence Investments and Finance Psychology Computer Science Algorithmic Trading Behavioral Finance Behavioral Analytics Sweet Spot
  13. 13. Application to Financial Intelligence 13 Customer Insight Human Resources Risk & Compliance Market Intelligence CRM Knowledge Extraction, Servicing & Cross-Sell Fraud/Misconduct Detection, Credit Scoring, Wall-Cross Recruiting/Retention, Talent Analytics, Automated Management Behavioral Investing, Alpha Generation, Liars Poker
  14. 14. Customer Insight •  Goal: leverage customer interactions to predict client preferences and behavior •  CRM, emails, documents, phone logs, chat, … 14 CRM Emails Voice Chat Knowledge/ Relationship Extraction Association Rule Learning
  15. 15. Natural Language Processing (NLP) 15 Intent   Sentiment Knowledge Graph Fact/Relation Extraction Named-Entity Recognition Syntactic Parsing Shallow Parsing POS/EOS Tokenization What’s  For   Dinner?  
  16. 16. Customer Insight Example •  Extracting Knowledge Base SVO triples from customer chat 16 *   *   Bid/Offer *   *   Financial Product Financial Product *   Bid/Offer *   *   *   Financial Entity Financial Benchmark Financial Benchmark Financial Benchmark Financial Model Financial Benchmark Bid/Offer *   Financial Level Financial Benchmark *  *   *   *   Subject Verb ObjectModifier Client buys 50M GM bonds
  17. 17. 17 Why  do  we  call  it  a  “Knowledge  Graph”?     Paul     Tudor  Jones   Michael     Cardillo   Galleon     Tech     Fund   Kris     Chellam   Krish     Panu   Ian     Horowitz   Stan     Druckenmiller   Leon     Shaulov   P&G     TransacHons   GS   TransacHons   Rajat  Gupta   Richard   SchuKe   Fantasy  Football     League   Wharton     School   Proctor  &     Gamble   Goldman     Sachs   SpotTail     Capital  Advisors   McKinsey     &  Co.   Intel  Corp.   Anil  Kumar   Gary     Cohn   David     Loeb   Lunch   Fund   Galleon   Group  LLC   Rajiv  Goel   P  &  G  Board   PorQolio   Manager   Raj  Rajaratnam   Voyager  Capital   Partners  
  18. 18. Customer Intelligence Graph •  Customer Touch Points •  Competitive Targeting •  Cross-Sell •  Up-Sell •  Orphans 18
  19. 19. Orphaned Customer 19 Help!!
  20. 20. Human Resources •  Goal: Talent recruitment, retention, placement, performance evaluation, promotion, compensation, workforce planning •  Focus: identifying qualities of high performers, changing outcomes (promote the best, encourage the rest) 20
  21. 21. Talent Analytics Example 21 •  Employee Retention: Real cost of losing a high-skill employee can often be > 1.5-2.0x annual salary Source: Dave Weisbeck
  22. 22. Risk & Compliance •  Goal: Financial, Legal, and Regulatory risk measurement and management •  Data Sources: internal communications (text, voice), client interactions, social media, HR, CRM, expenses, etc •  Event and Behavioral Anomaly Detection •  AML •  SAR •  Unauthorized Trading •  Manipulation •  Misconduct 22
  23. 23. Time Series Outlier Detection Box-Jenkins (ARIMA – autocorrelation) •  Pros: seasonality, stationarity detection •  Cons: subjectivity in parameter initialization Holt-Winters (Exponential Smoothing) •  Pros: seasonality (single), computation •  Cons: subjectivity in parameter initialization Seasonal Hybrid (Generalized) ESD •  Pros: seasonality (LOESS), local & global anomalies •  Cons: low recall 23
  24. 24. Example: Seasonal Hybrid ESD 24 Source:  
  25. 25. Relationship Graph 25
  26. 26. Market Intelligence •  Goal: Identifying and modifying individual and collective investor behavior for individual or collective gain •  Data Sources: news, SEC filings, corporate releases, public and private databases, electronic and written communications, audio/video, social media, Bud Fox 11/6/15 26 GekkoisticAltruistic
  27. 27. Behavior Modification •  After watching which movie would investors express greater risk tolerance in portfolio activities? 11/6/15 27 Fischer, Risk Analysis, Vol. 31, No. 5, 2011
  28. 28. Investor Paradox A fair coin is flipped: •  If heads, your net worth increases 40% •  If tails, your net worth decreases by 30% Would you play this game? How many times would you play? 11/6/15 28
  29. 29. Cognitive Error •  Game has positive expectation over short-term, but negative expectation over the long run 11/6/15 29 0 0.5 1 1.5 2 2.5 1 11 21 31 41 NetWorth Games Played Answer: Play Just Once
  30. 30. Which Would You Prefer? A.  $1 today? B.  $1.05 a month from now? What if the choices were: A.  $1 in 24 months from now? B.  $1.05 in 25 months from now? Most people prefer A in the first instance, B in second. Example of “hyperbolic discounting” 30
  31. 31. Cognitive Heuristics and Biases 11/6/15 31 Definition: a systematic pattern of deviation from norm or rationality in judgment, a mental shortcut where inferences are drawn in illogical fashion.
  32. 32. Example: Alpha Generation •  Goal: identify persistent sources of alpha •  Key is “persistent” as most signals (e.g. sentiment) are viewed as transitory and unstable over long-term •  Current approaches incorporate emotional, affect psychology and intent models that aggregate individual sentiment to identify short-term group trends •  Incorporate models of investor behavior? 32 Alpha: the active return above a benchmark representing investment skill
  33. 33. Sentiment Analysis “The outlook reflects our expectations that Apple will continue to maintain and defend its strong market position in mobile devices and tablets despite seeing an decline in its addressable market.” 33 Source: Yahoo
  34. 34. Behavioral Economics Traditional •  Perfect rationality •  Expected utility theory •  Risk aversion •  Efficient market hypothesis •  CAPM and APT Behavioral •  Bounded rationality •  Prospect theory •  Loss aversion •  Adaptive market hypothesis •  Behavioral Asset Pricing •  Market anomalies 34 Empirical evidence diverges greatly from classic theory Why do market anomalies exist?
  35. 35. Identified Market Anomalies 35 Year Authors Anomaly 1968 Ball & Brown Earnings announcement effect 1975 Haugen & Heins Low volatility effect 1976 Rozeff & Kinney January effect 1977 Basu Low P/E effect 1981 Gibbons & Hess Monday effect 1981 Shiller Excess volatility effect 1981 Banz Size effect 1985 Rosenberg, Reid, Lanstein Value effect 1987 DeBondt & Thaler Over-reaction to bad news 1993 Jegadeesh & Titman Momentum investing
  36. 36. Drivers of Anomalies Cognitive Errors •  Conservatism •  Overweight Bayesian priors •  Confirmation •  Selection bias •  Gambler’s Fallacy •  Mean reversion •  Representativeness •  Overweight recent evidence •  Availability •  Familiarity premium Emotional Biases •  Loss aversion •  Pain > Gain •  Overconfidence •  Underestimate risk •  Endowment •  Sentimental value premium •  Status-Quo •  Regret-aversion •  Herd behavior 36 Anomalies are drivers of excess return (alpha)
  37. 37. Quantitative Behavioral Finance •  Behavioral asset pricing model: •  This is a linear factor model that includes behavioral bias sentiment factors •  Exposure to behavioral biases given by weights γi 37 E r[ ]= rf + βj j=1 k ∑ (rj −rf )+ γi i=1 m ∑ bi Expected Return Risk-free Rate Factor Beta Bias Sentiment Premium Risk Factor Premium
  38. 38. In Search of Alpha 38 1.  Context-based analysis to extract features of behavioral biases observed in investors 2.  Combine these features with structured data to formulate behavioral bias factors 3.  Use the factors to fit regression models for asset returns 4.  Construct factor portfolios to monetize market anomalies by overweight/underweight exposure to specific behavioral biases Alpha Factor Portfolios Behavioral Analytics Data
  39. 39. Key Takeaways •  We have too much data, not enough actionable insights •  Humans are prone to cognitive “mistakes” (some more than others) •  Behavioral analytics can identify and quantify individual and group cognitive behavior •  The financial industry can benefit from much of the behavioral analytics insights developed in other industries •  There is a lot more to do! 39
  40. 40. Final Words •  We at Digital Reasoning are actively exploring neural embedding and deep learning methods to enhance feature extraction and prediction of behavioral biases •  Synthesys Studio: •  Book (Q2 2016): Python Machine Learning by Example 40