abhinavbatra06/fire-index-prediction

Predicting Forest Fire in Algeria # Fire Weather Index (FWI) Prediction ## Project Overview This project predicts the **Fire Weather Index (FWI)**, a key indicator of wildfire risk, using meteorological data. The dataset consists of weather observations recorded from **June to September 2012** for 2 regions ,namely the Bejaia region located in the northeast of Algeria and the Sidi Bel-abbes region located in the northwest of Algeria.. Three **linear regression models** (Linear, Ridge, and Lasso) were applied, with **hyperparameter tuning** to improve performance. --- ## Live Web App You can access the live app here ## Screenshot ## Dataset The dataset includes the following features: ### Weather Data Observations: - **Temperature (°C):** 22 - 42 - **Relative Humidity (%):** 21 - 90 - **Wind Speed (km/h):** 6 - 29 - **Rainfall (mm):** 0 - 16.8 ### Fire Weather Index (FWI) System Components: - **Fine Fuel Moisture Code (FFMC):** 28.6 - 92.5 - **Duff Moisture Code (DMC):** 1.1 - 65.9 - **Drought Code (DC):** 7 - 220.4 - **Initial Spread Index (ISI):** 0 - 18.5 - **Buildup Index (BUI):** 1.1 - 68 - **Fire Weather Index (FWI) (Target Variable):** 0 - 31.1 ### Requirements & How to Use - Install Dependencies: pip install -r requirements.txt - Clone the repository: - git clone github.com - cd fire-weather-index-prediction - jupyter notebook ## Approach #### Exploratory Data Analysis (EDA) - Data Cleaning - Removed null values and corrected column names. - Standardized categorical labels and converted region info into a binary column (is_sidi_bel_region) #### Feature Engineering - Added is_august as a seasonality feature (August had the highest fire occurrences). - Removed highly correlated features (BUI & DC) to reduce multicollinearity. #### Model Training - Train-Test Split: 80-20 split. - Regression Models Used: - Linear Regression (Baseline) - Ridge Regression (Best alpha = 5, L2 regularization) - Lasso Regression (Best alpha = 0.01, L1 regu …