Chiang Mai Journal of Science

     Species distribution models are a powerful tool to predict suitability map addressing ecology and conservation, especially of rare species. However, the limited occurrence data often decrease the performances of the prediction models. In this research, the Random Forest with Fuzzy selection of pseudo absence point (RFFA) method was created for habitat prediction of species with limited distribution data. In our study, Musa gracilis Holttum is naturally found only in Narathiwat, one of the southernmost provinces in Thailand. With only three collected localities, the species was used as a sample to test effi cacy of the RFFA method. The comparing the model results with real data, the statistical relationship, and the feasibility assessment of the two species distribution models. MaxEnt and RFFA methods showed that the performance of the RFFA model did not differ signifi cantly from that of MaxEnt in terms of effi ciency. It can be concluded from the model using the three-occurrence data that M. gracilis distributes in approximately 7,000 square kilometers, with limited boundary in Thailand peninsular and is facing a risk of extinction in the wild.