Run demo-1b43591076 on credit-g: planner=scripted, llm_mode=none (scripted, no LLM)

Data story

1000 rows, 21 columns.

Leakage findings

severitytypecolumns
none

Leaderboard

Validation roc_aucrf0.7997linear0.7913flaml0.7894hgb0.7753hgb0.7667dummy0.5

Top features (permutation importance, held-out slice)

Score drop when shuffled (roc_auc)checking_status0.08549credit_history0.02717credit_amount0.01953savings_status0.01717duration0.01443purpose0.01313installment_commitment0.01204foreign_worker0.009727other_payment_plans0.008635personal_status0.008305

Reviewer: APPROVE

Step latency

Seconds per stepprofile_dataset0.094detect_leakage18.05decide_leakage0train_model0.172train_model0.484train_model0.922train_model3.734tune_model2.015train_model16.02compare_models0.047review0explain_model1.235

Model card

# Model card: linear-ec8aac73a3

- Task: binary, metric roc_auc, target `bad_credit`
- Family: linear; validation roc_auc: 0.7913 (seeded holdout of the training file)
- Dropped columns (leakage/ID policy): none
- Unresolved leakage findings: 0
- Planner: scripted; LLM mode: none; seed 0
- Intended use: tabular prediction on data like the training file. Not validated for distribution shift. Holdout scores are optimistic relative to the final test score.