← Back
🏆

Kaggle Series

Competitive data science — notebooks, pipelines & reproducible ML. kaggle.com/wangdongyang

5+
Competitions
3080 Ti
GPU
0.947
Best LB
LGB+XGB
Stack

Active

⚡

Playground Series S6E5

Tabular binary classification — race lap performance prediction with GPU-accelerated ensemble.
LB 0.947 LGB + XGB Optuna 200t
Active · Deadline May 31
→

⚙️ Pipeline

📊

Baseline First

LGB/XGB baseline to establish a competitive floor before feature engineering.

🧪

Feature Engineering

Target encoding, interaction features, external data integration — CV-validated.

🎯

Optuna Tuning

200-trial TPE search across LGB + XGB parameter space, GPU-accelerated on 3080 Ti.

🔗

Ensemble Stacking

5-fold OOF blending with optimal probability calibration.