Fire risk prediction is critical for mitigating the devastating environmental and socio-economic impacts of wildfires. In this study, one of the machine learning methods was utilized for Fars Province, Iran, based on remote sensing data and geospatial (ground measurement) data. Noticeably, twelve important variables were considered, including physiographic variables (elevation, slope, aspect), climatic variables (max temperature, precipitation, humidity), human activity factors (distance to roads, settlements, rivers, agricultural lands), and vegetation variables (land use, vegetation type). The model was developed specifically for the summer/peak fire season and it was trained and validated on dataset of 1,089 historical fire incident points from 2019 to 2022, with sentinel-2 imagery used for visual validation of burned areas. The Random Forest model illustrated high ability and performance according to assessment section results; the Kappa coefficient was about 83 percent and important factors in this process included land use, vegetation type, and absolute maximum summer temperature. The findings indicated that fire occurrence is most strongly associated with the fourth class of elevation (1834–2289 m), the second class of slope, and, for distance maps, the first class of road map, settlement, river, and agricultural lands. This highlights the significant role of human ignition sources and specific environmental conditions. The fire risk map effectively categorizes the province into areas of differing vulnerability, offering a comprehensive, adaptable, and practical resource for land managers. This facilitates the formulation of tailored preventative strategies, optimal resource allocation, and a proactive methodology for wildfire management in diverse and vulnerable ecosystems.