Quick baseline and importance caveats

Fit a fast baseline (logistic regression / gradient boosting for A, linear regression for B) and look at feature importance as evidence, not truth. Concepts: Baseline model; permutation importance; caveats Project A: Logistic regression (class_weight='balanced') baseline: report Bad recall, Bad precision, ROC-AUC and business cost 5·FN + FP on the test split. Project B: Compare your baseline MAE with the TSO 'total load forecast' benchmark (the target to beat).