rel-amazon: user churn (Classification)

This example now sources data from the relbench Python library, not the old S3 parquet mirror. The runner materializes the official rel-amazon tables locally through get_dataset("rel-amazon", download=True) before building the GraphReduce frame.

from relbench_amazon_user_churn import run_rel_amazon_user_churn

df, auc, n_features, materialized, target = run_rel_amazon_user_churn()
print(materialized)
print(target, auc, n_features, len(df))

Current implementation:

• examples/relbench_amazon_user_churn.py
• examples/relbench_amazon_common.py

Current RelBench-aligned dates:

• validation cut date: 2015-10-01
• holdout cut date: 2016-01-01
• lookback start: 2008-01-01

Latest Tested Result

Latest local run from tests/data/relbench/run_reports/relbench_results.json:

• test ROC AUC: 0.6983
• test average precision: 0.7539
• feature count: 30
• rows: train 415013, validation 409792, test 351885
• runtime: 108.4s (2026-04-09T22:06:22Z)

Interactive Runner

Save API URL Run rel-amazon user churn

Idle