Pick features and prepare them for distances

Select application-time features, log-transform skewed ones, encode a few ordinal categories, scale everything. Concepts: Feature choice without target; scale dominance; mixed data Project A: Exclude credit_risk, personal_status_sex, is_foreign_worker. Use log credit amount, duration, age, installment rate, existing credits, checking status (ordinal), savings (known levels), owns_house (housing == own). Project B: Optional: 24 hourly values per day, standardised per day.