This project presents a complete machine learning workflow to predict the salaries of professional baseball players based on performance, experience, and league-level metrics.
Using advanced regression models and hyperparameter optimization techniques, we aim to create robust and interpretable models for player valuation.
🧠 Focus: Data science for sports analytics with real-world applications.
- 📂 Source: Kaggle Dataset
- 🎯 Target:
Salary - 🧾 Features: Batting stats, player experience, league details, career totals
BattingAverageCareerBattingAvgExperienceLevel(Binned)
| Step | Description |
|---|---|
| 🔍 EDA | Missing value handling, outliers, distributions |
| 🧪 Feature Engineering | Custom metrics, binning, encoding |
| ⚙️ Preprocessing | One-hot encoding, scaling |
| 🤖 Modeling | Linear, Ridge, Lasso, ElasticNet, XGBoost, LightGBM, CatBoost |
| 🔁 Hyperparameter Tuning | RandomizedSearchCV & Optuna |
| 🧬 Ensembling | Voting and Stacking Regressors |
| 📉 Evaluation | R², RMSE, MAE + visualizations |
| Model | Type |
|---|---|
| Linear Regression | Baseline |
| Ridge / Lasso / ElasticNet | Regularized Linear Models |
| Random Forest | Bagging-based Ensemble |
| XGBoost / LightGBM | Gradient Boosting Models |
| CatBoost | Boosting with Categorical Support |
| Voting / Stacking | Ensemble Meta-Learning |
✅ Best Performance: CatBoost on test set based on R² score.
- R²: Coefficient of Determination
- RMSE: Root Mean Squared Error
- MAE: Mean Absolute Error
- Residual and prediction comparison plots included.
You can run the full notebook on Kaggle:
İbrahim Ongören
📧 ongorenibrahim78@gmail.com
🌐 LinkedIn
This project is licensed under the MIT License – see the LICENSE file for details.
- Kaggle dataset contributor creepycrap
- Open-source Python community
- ML libraries:
Scikit-Learn,XGBoost,LightGBM,CatBoost,Optuna