data-validationlisted

## Purpose Catch data problems before they waste training time or produce misleading models. ## When to use it Before every modeling pass, especially when working with new or updated data. ## Checklist - **Shape and type check**: confirm columns, dtypes, and row counts match expectations. - **Null distribution**: count and visualize missing values per column; decide on imputation strategy. - **Value ranges**: check min/max/unique values for each feature; flag out-of-domain values. - **Train/test distribution**: compare feature distributions across splits (KS test or visualization); flag drift. - **Target leakage scan**: check for time-based leakage, ID columns, future-looking features in training data. - **Class imbalance**: quantify target distribution; plan for stratification or weighting if needed. - **Temporal ordering**: for time-series data, confirm no future data leaks into training windows. ## Expected output A data validation report with PASS/FAIL per check, attached to the experiment tracker entry.