Comparative Study of Classification Algorithms to Classify the Restoration Base on Burn Severity Level

Abstract

The land fire that occurred in September 2019 in Katingan Regency, Central Kalimantan Province, covering an area of 970.44 hectares, caused pollution losses and environmental damage. Estimating the size damaged by fire is essential to identify areas that need restoration and implement an effective fire management plan. In this study, fire severity data were obtained from Landsat 8 OLI images from July 2019 to December 2019, which were then clipped and stacked as preprocessing and processed with the RdNBR extraction feature to identify the density of the Spatio-temporal hotspot based on the burn severity index. Furthermore, the feature extraction results are modeled using two classification algorithms, namely the Random Forest Classifier and the ANFIS algorithm. The comparison results show that the value of precision, recall, and accuracy for the ANFIS algorithm is higher than the RF Classification. So it can be concluded that the ANFIS algorithm shows a more accurate performance than the RF Classification in classifying burned areas in Katingan Regency, Central Kalimantan.