applyInPandas / mapInPandas group-wise is useful when:

Question

Accepted Answer

You apply pandas/sklearn logic per group (e.g., per-store models) in a distributed fashion — Grouped pandas function API = per-store/per-entity sklearn in parallel.

Answer

You need to avoid Python entirely

Answer

Data has no groups

Answer

You want to train a global Spark ML Pipeline only

applyInPandas / mapInPandas group-wise is useful when:

Short Explanation and Infographic

Full Explanation