Monisha-NEU/Algerian-Forest-Fire-prediction

# Algerian Forest Fires Prediction This project analyzes the **Algerian Forest Fires dataset** to predict and classify forest fire occurrences based on environmental and meteorological factors. The dataset includes data from two regions of Algeria—Bejaia and Sidi Bel-Abbes—collected over four months in 2012. Multiple machine learning models were used to develop a predictive framework, offering actionable insights into fire occurrences and their severities. --- ## Table of Contents - Dataset - Project Workflow - Models Used - Key Results - Conclusion - How to Run the Project - References --- ## Dataset The dataset was sourced from the **UC Irvine Machine Learning Repository** and contains: - **244 instances**: 122 each from Bejaia and Sidi Bel-Abbes regions. - **11 attributes**: Environmental and meteorological features. - **1 output attribute**: Classifies instances as "fire" or "not fire." --- ## Project Workflow The project followed a 7-step methodology: 1. **Import and Examine the Dataset**: Loaded and examined the dataset using Python's `pandas` library. 2. **Data Pre-processing**: Cleaned data by removing 2 records with missing values, correcting data types, and saving a new CSV file. 3. **Exploratory Data Analysis (EDA)**: - One-hot encoding for fire classification. - Visualizations (histograms, heatmaps, pie charts, box plots) to explore data patterns. 4. **Feature Scaling**: Standardized features to improve model performance. 5. **Split Data**: Divided into 75% training and 25% testing sets. 6. **Model Training**: - Used **Linear**, **Ridge**, **Lasso**, and **ElasticNet** regression models. - Applied **LassoCV** for 5-fold cross-validation to optimize hyperparameters. 7. **Model Evaluation**: Evaluated using **Mean Absolute Error (MAE)** and **R-squared (R²)** metrics. --- ## Models Used - **Linear Regression**: A baseline model for understanding linear relationships. - **Ridge Regression**: Reduces overfitting by penalizing large coefficients (L2 …