Missingness patterns and missing by target/segment

Check whether missingness depends on other columns or on the target. Concepts: MCAR / MAR / MNAR (intuition); missingness as signal Project A: Hidden 'unknown' categories: savings_bonds = 'unknown/ no savings account' (bad rate ~17.5%, LOWER than small savers) and property = 'unknown / no property'. Treat 'unknown' as its own category, never as the lowest ordinal level. Project B: Are missing load hours clustered (one outage) or scattered? Plot missingness by month.