MHP

Machine Learning — Predictive ML Patterns

Tip prediction with Databricks sklearn + MLflow, Snowflake ML Functions & Snowpark ML
Databricks sklearn
GradientBoostingRegressor + Pipeline
MLflow autolog: params, metrics, artifacts
Batch inference → Gold Delta table
Snowflake ML
ML.FORECAST / ANOMALY_DETECTION (SQL)
Snowpark ML: sklearn API, data stays in SF
Feature Store + Container Runtime
dbt Feature Tables
Stable schema from Silver for ML
dbt test: quality gates before training
Lineage + version control in Git

Data leakage — the #1 ML mistake: never include target-derived columns (e.g. total_amount includes tip) as features