Data Science for Financial Inclusion
summary
Trusting Social replaces bank histories with telco call/SMS/top-up logs and combines other alternative sources to score the unbanked, claiming coverage of whole populations and better accuracy than financial-only FICO-style scores. Labels are sparse (defaults ~5–6% in examples), data runs to terabytes for 50M+ subscribers and scales toward 250M, and pipelines create ~10,000 features. Workflows use heavy feature engineering, graph analytics and representation learning, plus transformers, CV/NLP, KYC face recognition and chatbots; models run in production under continual revalidation.
AI-written summary · about
chapters (16)
- 0:00Introduction
- 1:10Outline
- 3:02Statistics
- 3:37Risk Scoring Model
- 4:25What do they need
- 4:57Problem
- 7:15Solution
- 11:47Modeling
- 14:25What do we need
- 14:57Challenges
- 21:34Infrastructure
- 22:12Scalability
- 23:10Reliability
- 24:15Challenge
- 25:46Summary
- 28:40Team
related talks
41mAI For All: Enhancing Digital Financial Services using Data Science
22mUsing data to innovate credit decisioning for Small-to-Medium Enterprises
01h 03mAI Innovation for a Smarter Financial Ecosystem
39mMachine Learning in Financial Credit Risk Assessment
27mData Science and ML with an emphasis on Banking...
36mFederated AI in Future Digital Banking
26mData Science is Key to Achieving Energy Access in Africa
35mBank evolution to data science-driven business model