Himanshutyagi4348/algerian_forest_fire_prediction

🔥 Algerian Forest Fire Prediction This project analyzes the Algerian Forest Fire dataset with a complete pipeline of EDA, feature engineering, visualization, and predictive modeling. The main focus is on training and tuning regression models to forecast forest fire occurrence. 📌 Project Overview Goal: Predict fire occurrence using regression models with optimized parameters Tech Stack: Python, Pandas, NumPy, Matplotlib, Seaborn, Scikit-learn Models Used: Linear Regression (with cross-validation) Ridge Regression (with RidgeCV) Elastic Net (with ElasticNetCV) 🛠️ Steps Performed 1. Data Cleaning & Preprocessing Removed missing values & inconsistencies Encoded categorical variables Scaled numerical features 2. Exploratory Data Analysis (EDA) Seasonal and regional fire trend analysis Correlation heatmaps & distribution plots Multicollinearity check (VIF analysis) 3. Feature Engineering Created new meaningful features Selected optimal subset of features for regression models 4. Data Visualization Plots with Matplotlib & Seaborn (scatter plots, boxplots, pair plots, heatmaps) Highlighted seasonal fire patterns across Algeria 5. Predictive Modeling & Hyperparameter Tuning LinearRegressionCV – optimized with k-fold cross-validation RidgeCV – tuned alpha values automatically ElasticNetCV – optimized alpha & l1_ratio with cross-validation Model evaluation metrics: R² Score MAE (Mean Absolute Error) MSE (Mean Squared Error) RMSE (Root Mean Squared Error) ├── data/ # Dataset files ├── notebooks/ # Jupyter notebooks (EDA, Modeling, Results) ├── src/ # Modular Python scripts │ ├── Ridge,Lasso__Regression.ipynb │ ├── model_training.ipynb ├── results/ # Visualizations & model metrics ├── README.md # Project documentation └── requirements.txt # Dependencies 📌 Future Work Test tree-based models (Random Forest, XGBoost, LightGBM) for non-linear relationships Deploy as a Streamlit/F …