Closing the Machine Learning Loop using Python
summary
A production pipeline written in Python + scikit-learn collects site events into an indicator table, triggers classification after 7 events critical mass, and pushes segment lookups for 3,240 precomputed buckets; a random forest reaches about random forest ~79% accuracy, ensembles nudge that further, and once inferred probability exceeds 88% threshold the profile moves to passive tracking while daily re-evaluation retrains models.
AI-written summary · about
chapters (12)
- 0:00Closing ML loop in Python
- 3:11What is ML and why it matters
- 8:31Designing feedback loops
- 13:45Data collection and storage
- 20:57Prototyping ML models
- 25:17Model accuracy and performance
- 32:24Avoiding data tracking pitfalls
- 37:48Training and testing in production
- 40:34ML with ad platforms
- 45:45Inferred data and human insights
- 46:06Live demo: gender inference
- 48:49Closing remarks and Q&A
related talks
37mAutomated Machine Learning: Building a Pipeline
25mML model from an idea to production with the help...
27mUsing Machine Learning to solve a classification problem with scikit-learn
42mMachine Learning Crash Course
01h 21mBuilding a ML Model Observability pipeline
35mAn End-to-End Machine Learning Ecosystem in a Quarter
33mRunning Machine Learning in Production - a Journey to Success
30mReconceptualizing Machine Learning for the Real-time World