Machine Learning on IBM i with scikit-learn

The question in 2019: can you run a scikit-learn model trained on IBM i Db2 data and write predictions back — all in one loop? Yes. This notebook does exactly that: connect via ibm_db, extract training data, train a classifier, score it, insert results back to IBM i.

IBM i shops sit on decades of clean, high-integrity business data — the exact kind of data ML models need and rarely get. The challenge was never the data. It was the mental model that IBM i couldn’t participate in the AI stack.

This project was the first step in what became: H2O Driverless AI on POWER9, Watson Studio + WML with IBM i, and eventually the ibmi-assistant LLM prototype. The thread runs all the way to IBM Bob.