Explainable AI Explained for Developers
summary
Demonstrates creating a churn classifier with Azure AutoML on ~10k–20k rows, producing a touted 99% model but showing that accuracy alone is insufficient; a web dashboard delivers prediction lists plus feature importance (permutation) and per-customer local explanations while population global explanations reveal delivery quality and recent rating as top drivers. Shows LIME-style text and image highlights, warns that explanations are another uncertain model, and uses segmentation to turn noisy local reasons into actionable group interventions.
AI-written summary · about
chapters (11)
- 0:00Intro
- 2:34Explain ML models
- 2:45AutoML
- 6:10Demo
- 10:37AutoML (continued)
- 14:22Why explainable AI?
- 19:58Demo
- 20:54Why explainable AI? (continued)
- 25:10Segmentation
- 30:42Takeaways
- 31:13Outro
related talks
47mML: Explainable Artificial Intelligence explained
16mBuilding Trust through Explainable AI
15mIntroduction to Explainable AI
58mDeep Dive into Azure Automated Machine Learning
24mWhy and How to Build Explainability into your ML Workflow
24mMachine Learning for Analytics Team
39mExplainable AI as a Conventional Data Analysis Tool
39mGuide towards algorithm explainability in machine learning