This project uses machine learning to classify fire types in India using MODIS satellite data from 2021 to 2023. The goal is to assist in real-time fire detection and improve disaster management strategies.
To build an accurate and efficient classification model that can identify the type of fire (e.g., forest, agricultural) based on satellite features and support rapid response systems.
- Python 3.10
- Pandas, NumPy, Scikit-learn
- Seaborn, Matplotlib
- SMOTE (imbalanced-learn)
- Joblib for model saving
- Data Collection from MODIS (2021–2023)
- Data Cleaning and Preprocessing
- Exploratory Data Analysis (EDA)
- Feature Selection and Encoding
- SMOTE for class balancing
- Model Training (Random Forest, KNN, etc.)
- Model Evaluation and Final Selection
- Model Saving with Joblib
Manual identification of fire types from satellite data is slow and error-prone. There is a need for a machine learning model that can automate fire classification to support faster decision-making during emergencies.
This project automates fire type classification using a Random Forest model trained on processed MODIS satellite data. The model uses key features like brightness, FRP, confidence, and temporal attributes to make predictions.
Due to GitHub's file size limit, the trained model (best_fire_detection_model.pkl) is stored on Google Drive.
🔗 Click here to download the model
Please download and place it in the project folder before running the notebook.
- Final Model: Random Forest Classifier
- Accuracy: High accuracy on balanced test data
- Evaluation Metrics: Accuracy, Precision, Recall, Confusion Matrix
- Supports faster identification of fire incidents
- Reduces manual efforts in classification
- Can be extended into live fire monitoring systems
Lavanya Parashar
AICTE Internship Project – Machine Learning
B.Tech CSE