Assurance Scoring Using Machine Learning and Analytics to Reduce Risk in the Public Sector
summary
A Python assurance scoring pipeline claims to identify and auto-route the majority of low-risk public sector cases so scarce investigator time focuses on exceptions it implements feature engineering, cross‑validation, thresholded classifiers and monitoring with Pandas & Scikit-Learn then boosts confidence with deterministic rules, statistical anomaly detection and neo4j graph traversal probabilistic linking it targets extreme class imbalance around less than 0.1% fraud and reports reproducible pipelines, yearly holdout tests and deployed matching that cut manual reviews.
AI-written summary · about
chapters (14)
- 0:00Assurance scoring in public sector
- 2:16Problem statement and goals
- 4:07Olympics ticket fraud example
- 6:14Traditional fraud detection overview
- 8:24Assurance scoring pipeline
- 10:33Analytical data model with POLE
- 13:45ML classification framework
- 16:12Feature engineering
- 19:53Model selection and thresholds
- 21:59Refining buckets with rules
- 24:06Anomaly detection for screening
- 24:49Risk linkage with graph traversal
- 26:58Pipeline validation and deployment
- 27:38Q&A
related talks
08mAI-Powered Fraud Detection in Financial System
39mMachine Learning in Financial Credit Risk Assessment
22mMachine Learning–Driven Third-Party Risk Management at Scale
01h 30mAI in Finance From Insight to Action
18mTransforming Financial Decision Intelligence Through Advanced NNs
11mBuilding Trust in AI-Enhanced Public Sector Systems
01h 21mEmpirical methods fairer AI - Auditing a Dutch public sector risk profiling algorithm
47mKeynote: Using Data Science for Social Good: Examples, Opportunities, and Challenges