Experimental Exploratory of Temporal Sampling Forest in Forest Fire Regression and Classification

Abstract

Temporal Sampling Forest (TS-F) has been devoted to tackle the sequential data classification problem. It extends the robustness of random forest (RF) in handling the sequential data classification. However, it has not been used in the area of forest fire detection. Forest fire can be seen as a temporal phenomenon where it does not form in one day, but subsequently occurred due to the sequential changes of climates, human factors, and other affecting factors. Therefore, this paper is aim to tackle the data of forest fire from two perspectives, which are regression analysis and classification problem by using TS-F. The root mean square error (RMSE) obtained for regression analysis in the 1st dataset is 61.35 while the classification accuracy of the 2nd dataset is 84.18 %.