Levi-acker07man/Algerian-Forest-Fire-Model-Training-Project

# 🌲 Algerian Forest Fire Prediction Pipeline ### Tech Stack Used ## 📌 Project Overview This is an end-to-end Machine Learning regression project designed to predict the **Fire Weather Index (FWI)** of the Algerian forest regions based on meteorological data. The project encompasses the entire data science lifecycle: taking raw, unformatted data, structurally cleaning it, performing extensive Exploratory Data Analysis (EDA), building a mathematically robust regularized model to avoid overfitting, and finally deploying that model to a live **Flask Web Application** for real-time user predictions. ## 🧹 Phase 1: Data Cleaning & EDA Real-world data is inherently messy. A significant portion of this project was dedicated to strictly preprocessing the raw CSV before any machine learning could occur. * **Dataset Merging:** Combined instances from two distinct regions (Bejaia Region and Sidi Bel-abbes Region) into one unified DataFrame to increase the model's geographical generalization. * **Structural Formatting:** Stripped hidden whitespace from column headers and string values to prevent key errors during automated processing. * **Handling Missing Data:** Programmatically identified and dropped corrupted rows and `NaN` values. * **Type Casting & Encoding:** Converted raw string data into workable numeric formats (integers and floats) and transformed the textual target column (`Classes` containing "fire" and "not fire") into binary numerical values (`1` and `0`) for mathematical correlation. ## 🧠 Phase 2: Model Training & Regularization To ensure the model performs perfectly on unseen testing data without simply memorizing the training dataset, standard Linear Regression was bypassed. * **Feature Scaling:** Applied `StandardScaler` to ensure all meteorological features (Temperature, Wind Speed, Humidity) contributed equally to the model without dominating one another due to differing units. * **Regularization:** Rigorously tested **Lasso Regression (L1)**, ** …