This project aims to accurately predict the demand for a bike-sharing service. By analyzing historical data, we built a machine learning model to forecast the number of bike rentals based on various factors like weather conditions, time of day, and seasonal information. The primary goal is to provide a predictive tool that can help optimize bike fleet management and ensure timely availability for users.
- Data Preprocessing: Handled missing values, formatted data, and engineered new features from existing data.
- Exploratory Data Analysis (EDA): Explored the relationships between weather, time, and bike rental counts.
- Model Building: Utilized a powerful XGBoost Regressor model for its high performance and robustness.
- Hyperparameter Tuning: Employed Grid Search with cross-validation to find the optimal model settings and prevent overfitting.
- Model Evaluation: Evaluated the model's performance using industry-standard regression metrics (R-squared, MAE, RMSE).
The final tuned model achieved outstanding results on the unseen test data:
- Mean Absolute Error (MAE): ~34.79
- Root Mean Squared Error (RMSE): ~53.81
-
R-squared (
$R^2$ ): ~0.91
- Python
- Pandas
- Numpy
- Scikit-learn
- XGBoost
- Matplotlib & Seaborn
- MonarchofCoding