Forest fires have been detected to occur in Indonesia since 1998. The forest fires mostly resulted from human activities in order to expand their land, especially oil palm lands. Once a land clearing is carried out, it covers areas of tens to hundreds of thousands of hectares. The peak of the forest fires occurred in 2015 when almost half of the world was affected especially by smoke from the fire. In 2011, VIIRS (Visible Infrared Imaging Radiometer Suite) was launched by NOAA (National Oceanic and Atmospheric Administration). In their collected data, there is data on Indonesia issued daily for areas in the islands of Java, Sumatra, Kalimantan, Sulawesi, and the general areas of Indonesia. The VIIRS night data shows the possibility of some burning at night. We need to identify burnings that potentially become fires, which are potentially producing smoke. It is important to observe areas where large-scale forest fires frequently occur, so the authority could do some prevention beforehand. Using existing data released per day per area, we utilize an ANN (Artificial Neural Network) to identify a potential fire. The processes start with data cleaning and processing, Neural Network creation, and finally ANN training and testing. By using the ANN prediction model with the MinMax Scaler, we choose variables Temperature, Radiant Heat Intensity, and Source Footprint. Simulations are made to show the ranges of these variables that predict the possibility of 'flame' status occurring. We conclude that the ANN method can give more accurate results than the existing classification.