How accurately can we predict a power plant's electrical output from changing environmental conditions?
In this project, I built and evaluated a supervised machine learning pipeline to predict electrical power output (PE) for a combined-cycle power plant using four environmental measurements: ambient temperature, exhaust vacuum, ambient pressure, and relative humidity.
The project follows a complete end-to-end machine learning workflow, including exploratory data analysis, feature selection, train-test splitting, cross-validation, model comparison, hyperparameter tuning, and final model evaluation. Three regression algorithms—Linear Regression, Ridge Regression, and Random Forest Regression—were compared using 5-fold cross-validation to identify the best-performing model.
After tuning with GridSearchCV, the Random Forest model achieved the strongest predictive performance, explaining approximately 96.4% of the variance in power output (R² = 0.964) while maintaining low prediction error.
This project demonstrates practical experience with the full machine learning modeling process—from selecting appropriate algorithms to validating performance and interpreting results—using Python and scikit-learn.
- Exploratory Data Analysis (EDA)
- Supervised Machine Learning (Regression)
- Feature Engineering
- Train/Test Splitting
- Cross-Validation
- Model Comparison
- Hyperparameter Tuning (GridSearchCV)
- Performance Evaluation (MAE, RMSE, R²)
- Python, Pandas, NumPy, scikit-learn