Detection of forest fire using support vector machine in comparison with random forest to measure accuracy, precision and recall

Machine learning techniques are widely used in forest fire detection due to its accurate results in detection. The main objective of this proposed work is to evaluate the performance of Novel Support Vector Machine in comparison with Random Forest algorithm for detection of Forest fire. A total no of 1599 samples are collected from the dataset available in UCI Repository. These samples are divided into 70 % for the training dataset (n = 1119) and 30 % for testing dataset (n = 480). Accuracy, Recall and Precision values are determined to evaluate the performance of the Novel Linear Support Vector Machine algorithm. Novel Linear Support vector machine achieved accuracy of 96 % whereas Random Forest algorithm achieved 78.50%. Precision obtained for the SVM algorithm is 94.04 % and 73.05 % for Random Forest algorithm. Recall obtained for SVM algorithm is 94.13 % and 78.52 % for Random Forest algorithm. The significant value achieved is 0.025 (p ! 0.05). In this study, it is observed that the SVM algorithm performed significantly better than the Random forest in detection of forest fire for the dataset examined.