Skip to content

Generating Test Predictions with Optimized Model

Model Inference

Using the champion XGBoost model (version 10) to generate competition predictions:

File: src/predict.py

python src/predict.py \
     --model-uri models:/bank-client-subscription-classifier-xgboost/10 \
     --model-type xgboost \
     --input-path data/processed/test_processed.csv \
     --output-path data/predictions/xgboost_predictions.csv

Competition Results

Optuna Competition Results V2

Performance Analysis

The Optuna-optimized XGBoost model showed improvements in the competition leaderboard:

  • Leaderboard Position: Improved from rank 1,427 to 1,193 (234 position improvement)
  • Competition Score: Enhanced from V1's baseline to 0.96878
  • Local Validation: AUC improved from 0.966 to 0.968

While this isn't a dramatic gain, it's still a solid improvement that validates the hyperparameter optimization approach.