Using Markov Chain approximation to identify financial fraud
summary
Claims that sequential web data over six years yields predictive features and that a Markov approach helps spot anomalies in production are backed by concrete steps: encoding changes into three-feature states, estimating a 27×27 state matrix, and extracting features from sequence probabilities. A demo notebook and a scikit‑learn style class fit on simulated data (about 10,000 entries 1,000 sequences), compute transition frequencies and log probabilities, and feed those features into a model that successfully separates fraudulent-like sequences from normal behavior.
AI-written summary · about
chapters (8)
- 0:00Markov chain fraud detection
- 1:05Financial fraud types
- 6:16Web data sequential patterns
- 9:11Encoding sequential data
- 11:46Anomaly detection with Markov chains
- 12:49Markov chain probabilities
- 17:09Demo notebook
- 23:49Conclusions and future
related talks
16mLeveraging ML to Create a Reliable Anti-Fraud System
52mModern Fraud Prevention Using Deep Learning
24mMachine learning and financial time series analysis
40mFraud detection without labeled data
01h 30mAI in Finance From Insight to Action
44mIntroduction to Sequence Analytics
23mWhy Fraud Detection, Sports Betting, and Bitcoin Prediction Look Alike
13mBehavioral Analytics for Real-Time Fraud Prevention