Data Science and ML with an emphasis on Banking...
summary
Banking use case predicts loan default from a dataset with 88% non-defaulters, using support vector machines, random forest, and extreme gradient boosting and addressing class imbalance via under-sampling. Before balancing models show ~85–88% accuracy but only ~31–35% sensitivity; after under-sampling accuracy shifts to ~78–81% while sensitivity rises to ~80–83%. Team deploys a real-time API and reports the new system detects defaulters about four times more effectively than the legacy approach.
AI-written summary · about
chapters (15)
- 0:00Start
- 1:24Data Science & Machine Learning
- 4:11Machine Learning Applications
- 7:53Challenges in Data Preprocessing
- 15:44Essential Workflows
- 17:30Classification in Machine Learning
- 19:11What is bank loan default
- 20:15Approaches to Handle Imbalanced Data
- 21:19Architecture of Proposed Model
- 21:37Correlation Plot
- 22:55Accuracy of Algorithms Before Addressing Class Imbalance Issue
- 23:40Sampling Approach: Under-sampling
- 24:23Accuracy of Algorithms After Addressing Class Imbalance issue by Using Under-sampling
- 25:16Use Case in Banking Industry
- 26:10Conclusion
related talks
39mMachine Learning in Financial Credit Risk Assessment
04h 03mTutorial - Intro to Data Science
35mBank evolution to data science-driven business model
23mChallenges in using machine learning for financial monitoring
58mNew-type Scoring Models: Analyzing Customer Actions
30mMachine Learning Techniques for Class Imbalances & Adversaries
31mData Science for Financial Inclusion
01h 39mApplying Rigorous Machine Learning Methods in Business Strategy