Bedant03/Algerian-Forest-fire-Prediction

# Algerian Forest Fire Prediction using Ridge Regression This project focuses on building a machine learning regression model to predict the Fire Weather Index (FWI) using meteorological and fire-related features from the Algerian Forest Fires dataset. The trained model is later serialized and deployed using a Flask web application for real-time prediction. Problem Statement: Forest fires are strongly influenced by environmental conditions such as temperature, humidity, wind speed, and moisture indices. The goal of this project is to predict the FWI (Fire Weather Index), which represents the intensity and spread potential of fires, using historical weather and fire data. This is formulated as a supervised regression problem. Dataset Description The dataset contains 243 records with meteorological, fire index, and regional information collected from two regions in Algeria. After preprocessing, the main features used include temperature, relative humidity, wind speed, rainfall, FFMC, DMC, DC, ISI, BUI, fire class, and region. The target variable is FWI (Fire Weather Index). Data Cleaning and Preprocessing Initial inspection showed redundant date-related columns (day, month, year), which were removed as they did not add predictive value. The Classes column contained inconsistent string labels such as "fire" and "not fire" with spacing issues. This column was cleaned and converted into a binary numerical format where: 1->represents fire 0->represents no fire Correlation analysis was performed using a heatmap. Highly correlated features (correlation > 0.85) were identified, and the feature DC was removed to reduce multicollinearity. All numerical features were standardized using StandardScaler to ensure stable model training. Model Training Multiple regression models were evaluated: 1.Linear Regression 2.Ridge Regression 3.Lasso Regression 4.ElasticNet Regression The dataset was split into training and testing sets using a 75–25 split. Ridge Regres …