Forests are the lungs of the world as a valuable supporter of human health. Forest destruction can cause damage to the world's air circulation. One of the causes of forest destruction is wildfires (Ardiyanto and Hidayat, 2020). Wildfires have recently attracted increasing international attention as an environmental and economic issue, and are considered a potential threat to the survival of living things. Wildfires can be caused by both natural and man-made factors. Natural factors include natural disasters and gusts. Man-made factors include human carelessness and deliberate forest burning (Ramli et al., 2021). Wildfires in Indonesia are a high-intensity and recurring problem. Wildfires start with small hotspots, then grow larger and larger as conditions in the field change. Anticipatory measures are efforts to spread the fire widely, so that the number of losses and negative impacts can be minimised. One of the efforts to anticipate the spread of fire is to investigate the type of fire early on so that it can be handled early on. The process of identifying hotspots in wildfires can be done with a hotspot classification approach as a preventive measure against the spread of burning land (Dwiasnati and Devianto, 2021). There have been many studies that implement classification algorithms to detect/investigate the types of fires in wildfires in various regions. A study in (Dwiasnati and Devianto, 2021) used Machine Learning algorithms (Naïve Bayes, SVM, and K-Nearest Neighbour (K-NN)) to estimate the area of wildfires in the Kampar region, Riau. In his research, the K-NN algorithm provides the best accuracy compared to other algorithms. Research in (Pratiwi et al., 2018) classified forest and land fires in Pelalawan Regency, Riau using the Naïve Bayes algorithm. The attributes used for classification consist of temperature, humidity, rainfall and wind speed. The Naïve Bayes algorithm gave an accuracy result of 82 per cent. Research (Khairani and Sutoyo, 2020) revealed that the Random Forest algorithm is the best algorithm for classifying hotspots in Kalimantan. The attributes used as parameters are climatological information from BMKG. Previous research and using the same data was conducted by (Karo, 2020), they identified the type of hotspots using the XGBoost and Feature Importance algorithms. Feature Importance is a feature selection method used. The results of his research revealed that the combination of XGBoost and Feature Importance was superior to the SVM, decision tree and logistic regression algorithms.

In this research, efforts to identify the type of fire by classifying fire points. The dataset used is sourced from Global Forest Watch (GFW), which has also been implemented in previous studies. The classification algorithms used include K-Nearest Neighbour (K-NN), Naïve Bayes and Iterative Dichotomicer 3 (ID3) combined with feature selection.