Prepare a small labeled dataset and confirm how attributes are interpreted before training. Choose an evaluation procedure that reflects the real prediction task, and prevent information from the test set entering preprocessing. Record model settings and compare a simple baseline before increasing algorithm complexity.
This is best suited to teaching and experimenting with tabular machine learning. Before committing a larger project, account for these trade-offs: interface expects machine-learning concepts; in-memory workloads can be constrained by RAM.
