Enhancing Forest Fire Prognosticating Using Machine Learning Algorithms

Abstract

Introduces a machine learning-based approach that uses certain geographic and environmental characteristics to categorize different types of forest cover. The Random Forest Classifier technique, which effectively manages classification problems by building many decision trees and combining their output for increased accuracy and robustness, is used to design the model. Important input variables including height, aspect, slope, distances to roads and hydrology, hill shade values, and particular soil and wilderness kinds are all taken into account by the algorithm. To generate predictions, users can manually enter numbers or upload datasets using an intuitive interface created using Streamlit. Apart from producing forecasts, the system offers assessment metrics like accuracy, precision, recall, and F1 score, which are bolstered by visual aids like metric bar charts and confusion matrix plots. Through a dependable and understandable machine learning model, this structured prediction approach provides insights for applications involving the analysis and classification of forest data.

Introduction

1. Objective and Approach

The research aims to classify forest cover types and predict forest fires using machine learning, specifically the Random Forest Classifier (RFC). It utilizes structured data with topographical and environmental features such as:

• Elevation, slope, aspect

• Soil type

• Proximity to roads or hydrology

• Wilderness area classification

By integrating statistical techniques and algorithmic learning, the system detects patterns in labeled datasets for accurate classification and forecasting.

2. Key Components

A. Random Forest Classifier

• An ensemble model aggregating decisions from multiple decision trees.

• Uses random subsets of data and features to prevent overfitting.

• Offers high accuracy, robustness to noise, and feature importance insights.

B. Forest Cover Type Prediction

• Predicts types of forest land using spatial and environmental data.

• Involves both numerical (e.g., elevation) and categorical (e.g., soil type) features.

• Helps in landscape analysis, resource planning, and ecological studies.

C. Machine Learning in Classification

• Focuses on supervised learning for classification tasks.

• Utilizes algorithms like Decision Trees, SVM, and Random Forest.

• Features are crucial for training effective models, and feature engineering enhances performance.

3. Literature Review Highlights

Several models were studied for forest fire prediction, often integrating GIS, satellite data, and ensemble methods. Key findings include:

• FFSBP Model: Combines fire spread direction and burned area prediction, showing high performance in China and Portugal.

• India (NE Region): Fires are mostly human-caused and seasonal; geospatial tools and community engagement are recommended.

• Parallel SVM: Optimized for large datasets using Apache Spark.

• NASA MODIS & RF: Utilized for high-risk area prediction with real-time alerts.

• Central-South China: LightGBM with GIS data accurately predicts fire-prone months and regions.

• U.S. Forests: RFC outperformed other models using meteorological factors, achieving 86.5% accuracy.

• Ensemble Methods (e.g., DNN, XGBoost, SHAP, LIME): Help improve accuracy and explainability in models.

• NE India: Random Forest showed AUC = 0.87, identifying key fire-driving factors like vegetation and human density.

4. Existing System

• Predicts forest fires using environmental metrics (temperature, humidity, wind, etc.).

• Focuses on early detection and real-time alerts.

• Aims to reduce the ecological and human impacts of wildfires using automated pattern detection.

5. Proposed System

Architecture & Design

Uses a Random Forest model for forest cover classification based on structured input. The system includes:

• A. Data Input:

  • Manual or file-based (CSV) input

• B. Pre-Processing:

  • Scales numerical data, encodes categoricals, handles missing values

• C. Prediction Module:

  • Uses trained RFC to classify cover type

  • Results shown on-screen and downloadable

• D. Model Evaluation:

  • Metrics: Accuracy, Precision, Recall, F1 Score

  • Confusion matrix for performance visualization

6. Visualization

• Heatmaps (confusion matrix) and bar charts illustrate model performance.

• Helps interpret prediction quality across different classes.

7. Result Analysis

• Model evaluation uses:

  • Accuracy: Overall correctness

  • Precision: Accuracy of positive predictions

  • Recall: Ability to identify all relevant cases

  • F1 Score: Balance between precision and recall

These metrics provide a comprehensive view of the classifier’s effectiveness, enabling both technical validation and practical deployment.

Conclusion

To sum up, the forest cover type prediction system shows promise in efficiently categorizing different types of cover according to specific environmental characteristics. In addition to making predictions, the system uses a machine learning model and integrates assessment metrics including accuracy, precision, recall, and F1 score to assess the model\'s performance. Together with the confusion matrix and visualizations, the evaluation results provide insightful information about the model\'s advantages and shortcomings. The model can be improved to produce predictions with greater accuracy and dependability by comprehending these criteria. In order to keep the model efficient and flexible for upcoming use cases, this method emphasizes the significance of ongoing development and optimization in machine learning systems.

References

[1] Ananya Shahde, Aditi Shahdeo, Prakruthi S Reddy, Chaitra K, Wildfire Prediction and Detection using Random Forest and Different Color Models, International Research Journal of Engineering and Technology (IRJET), Volume: 07 | June 2020 [2] According to Nhu, V.-H., Janizadeh, S., Avand, M., Chen, W., Farzin, M., Omidvar, E., Shirzadi, A., Shahabi, H., Clague, J.J., Jaafari, A., et al. A Comparison of Computational Ensemble Data Mining Models for GIS-Based Gully Erosion Susceptibility Mapping. Appl. Sci. 2020, 10, 2039. [Reference] [3] Arunkumar, Kanimozhi, Forest Fire Prediction Using Machine Learning Techniques, International Journal of Scientific Research in Engineering and Management (IJSREM). [4] Bui, D.T.; Khosravi, K.; Tiefenbacher, J.; Nguyen, H.; Kazakis, N. Using innovative hybrid machine-learning techniques to improve water quality index prediction. Environ. Sci. Total. 2020, 721, 137612. [Reference] [Med] [5] Deepenbacher, J.P.; Pradhan, B.; Tien Bui, D.; Arabameri, A.; Chen, W.; Blaschke, T. Water 2020, 12, 16. Gully head-cut distribution modeling with machine learning techniques—A case study in NW Iran. [6] Hoang Thi Hang, Javed Mallick, Saeed Alqadhi, Ahmed Ali Bindajam, Hazem Ghassan Abdo, Exploring forest fire susceptibility and management strategies in Western Himalaya: Integrating ensemble machine learning and explainable AI for accurate prediction and comprehensive analysis, Science direct, Volume 35, August 2024, 103655. [7] In Da Lat City, Lam Dong Province, Vietnam, Thanh, D.Q.; Nguyen, D.H.; Prakash, I.; Jaafari, A.; Nguyen, V.-T.; Van Phong, T.; Pham, B.T. employed a frequency ratio approach based on Gis for mapping landslide susceptibility. Earth Sci. Vietnam J. 2020, 42, 55–66. [8] Jakkamsetti Bharath, K. Swapna, Forest Fire Prediction Using Supervised Machine Learning Models, IJFANS INTERNATIONAL JOURNAL OF FOOD AND NUTRITIONAL SCIENCES, Volume 13, 2024. [9] J. Shi and J. Zhang, ‘‘Load forecasting based on multi-model by stacking ensemble learning,’’ Proc. CSEE, vol. 39, no. 14, pp. 4032–4042, 2019. [10] Kajol R Singh, K.P. Neethu, K Madhurekaa, A Harita, Pushpa Mohan, Parallel SVM model for forest fire prediction, Soft Computing Letters, 2021, Soft Computing Letters 3 (2021) 100014 [11] Li, S.; Hong, H.; Wang, X.; Bian, H.; Zhang, S.; Pradhan, B.; Li, Y.; Xue, W.; Shahabi, H.; Li, S. Chen, W. utilizing data-driven techniques such as random forest, alternating decision tree, and naïve bayes tree to model flood susceptibility. Environmental Science, 701, 134979, 2020. [12] M. Seddouki, M. Benayad, Z. Aamir, M. Tahiri, M. Maanan, and H. Rhinane, ‘‘Using machine learning coupled with remote sensing for forest fire susceptibility mapping. Case study Tetouan province, Northern Morocco,’’ Int. Arch. Photogramm., Remote Sens. Spatial Inf. Sci., vol. XLVIII-4/W6, pp. 333–342, Feb. 2023. [13] Nazimur Rahman Talukdar, Firoz Ahmad, Laxmi Goparaju, Parthankar Choudhury, Rakesh Arya, Abdul Qayum, Javed Rizvi, Forest fire estimation and risk prediction using multispectral satellite Images, Natural Hazards Research, 2024, Natural Hazards Research 4 (2024) 304–319. [14] Performance evaluation of machine learning algorithms for forest fire modeling and prediction, Symmetry 12 (6) (2020) 1022, B.T. Pham, A. Jaafari, M. Avand, N. Al-Ansari, T. Dinh Du, H.P.H. Yen, T.V. Phong, D.H. Nguyen, H.V. Le, D. Mafi-Gholami, I. Prakash. [15] Phong, T.V.; Ly, H.-B.; Le, T.-T.; Trinh, P.T.; Dao, D.V.; Jaafari, A.; Bayat, M.; Mafi-Gholami, D.; Qi, C.; Moayedi, H.; et al. A deep learning neural network model with spatial explicitness for landslide susceptibility prediction. 2020, Catena, 188, 104451. [16] Quansheng Hai, Xiufeng Han, Battsengel Vandansambuu, Yuhai Bao, Byambakhuu Gantumur, Sainbuyan Bayarsaikhan, Narantsetseg Chantsal and Hailian Sun, Predicting the Occurrence of Forest Fire in the Central-South Region of China, MDPI (Multidisciplinary Digital Publishing Institute) Forests 2024, 15, 844. https://doi.org/10.3390/f15050844. [17] R. Alkhatib, W. Sahwan, A. Alkhatieb, and B. Schütt, ‘‘A brief review of machine learning algorithms in forest fires science,’’ Appl. Sci., vol. 13, no. 14, p. 8275, Jul. 2023. [18] Shahabi, H.; Geertsema, M.; Khosravi, K.; Amini, A.; Bahrami, S.; Shirzadi, A.; Ghaderi, K.; Omidvar, E.; Al-Ansari, N.; Clague, J.J. Using machine learning and sentinel-1 remote sensing data for flood detection and susceptibility mapping: Bagging ensemble hybrid intelligence based on k-nearest neighbor classification. 2020; Remote Sens. 12, 266. [19] Shallow Landslide Susceptibility Mapping by Random Forest Base Classifier and its Ensembles in a Semi-Arid Region of Iran Nhu, V.-H.; Shirzadi, A.; Shahabi, H.; Chen, W.; Clague, J.J.; Geertsema, M.; Jaafari, A.; Avand, M.; Miraki, S.; Asl, D.T. 2020, 11, 421; Forests. [Reference] [20] Tiefenbacher, J.P.; Pradhan, B.; Jahani, A.; Goshtasb, H.; Kornejady, A.; Shahabi, H.; Rahmati, O.; Panahi, M.; Ghiasi, S.S.; Deo, R.C.; et al. Neural fuzzy hybrid ensembles for modeling and forecasting dust sources. 2020, Atmos. Environ. 224, 117320. [Reference] [21] Volume 07 | (2023), DOI: 10.55041/IJSREM19604. [22] Yashnil Mohanty, A Machine Learning Approach to Predict the Occurrence of Forest Fires with Meteorological Parameters, high school edition journal of student research, Volume 13 |(2024). [23] Z. Wang, K. Wang, Y. Li, and G. Li, ‘‘Research on forest fire prediction in Yunnan province based on LightGBM and SHAP,’’ Fire Science Technol., vol. 42, no. 11, pp. 1567–1571, 2023. [24] Zechuan Wu, Bin Wang, Mingze Li, Yuping Tian, Ying Quan, Jianyang Liu, Simulation of forest fire spread based on Artificial Intelligence, Ecological Indicators, 2022, Ecological Indicators 136 (2022) 108653. [25] Z. Zhou, Integrated Learning: Fundamentals and Algorithms. Beijing, China: Publishing House of Electronics Industry, 2020.

Copyright

Copyright © 2025 K. Mathivanan, G. Bavadharani, R. Nithya . This is an open access article distributed under the Creative Commons Attribution License, which permits unrestricted use, distribution, and reproduction in any medium, provided the original work is properly cited.