Nidhi18-git/algerian-forest-fire-fwi-prediction

# Algerian Forest Fire FWI Prediction An end-to-end machine learning project that explores Algerian forest fire data and predicts the **Fire Weather Index (FWI)** from weather observations and fire-danger indicators. The project covers data cleaning, exploratory data analysis (EDA), feature engineering, multicollinearity handling, model comparison, and serialization of the selected preprocessing and regression artifacts. ## Project overview Forest fire risk is strongly influenced by weather and fuel-moisture conditions. This project uses observations collected in 2012 from two Algerian regions—Bejaia and Sidi Bel-Abbes—to build a regression model for FWI. The workflow includes: - cleaning the original region-separated CSV data; - adding a numerical region indicator; - correcting column names and data types; - encoding fire and non-fire observations; - exploring feature distributions, class balance, correlations, and seasonal trends; - removing highly correlated predictors to reduce multicollinearity; - standardizing the input features; - comparing Linear, Lasso, Ridge, and Elastic Net regression models; and - saving a trained Ridge model and fitted scaler for reuse. ## Dataset The dataset contains daily observations from **June to September 2012** for two regions of Algeria: - **Bejaia**, in northeastern Algeria - **Sidi Bel-Abbes**, in northwestern Algeria The original dataset contains 244 observations (122 per region). After removing separator/header rows and one incomplete observation, the cleaned dataset contains **243 rows**. ### Features | Feature | Description | |---|---| | `Temperature` | Noon maximum temperature in °C | | `RH` | Relative humidity (%) | | `Ws` | Wind speed (km/h) | | `Rain` | Daily rainfall (mm) | | `FFMC` | Fine Fuel Moisture Code | | `DMC` | Duff Moisture Code | | `DC` | Drought Code | | `ISI` | Initial Spread Index | | `BUI` | Buildup Index | | `FWI` | Fire Weather Index (prediction target) | | `Classes` | Fire / not-fire label …