subtopic

End-to-end prediction function with input validation

Write predict_one(raw_input: dict) that validates inputs, builds features, applies the pipeline and returns a decision with reasons. Concepts: Inference pipeline; same transformations at train and serve time; validation Project A: Input: one loan application → P(Bad), decision at the cost threshold, top-3 SHAP reasons. Project B: Input: date + hour + temperature forecast → load forecast in MW (lags pulled from history).