Early identification of geological disasters and monitoring of easy-happening areas are important work in disaster prevention and reduction. In this paper, taking Pingyi county in Shandong province as the study area, the GF 1 WFV optical image, ASTER GDEM terrain data and precipitation data are fused into multi-source heterogeneous data. The extraction effects of three machine learning algorithms, such as TensorFlow algorithm, support vector machine, and random forest in geological hazard easy-happening areas have been compared. Geological hazard easy-happening areas in the study area from 2021 to 2024 have been extracted. By using TensorFlow algorithm, support vector machine and random forest methods, landslide easy-happening areas can all identified well. compared to other methods, TensorFlow algorithm has a higher classification accuracy with an overall accuracy of 82.33% and a Kappa coefficient of 0.82. From 2021 to 2024, the proportion of geological hazard easy-happening areas in Pingyi county ranged from 11.5% to12.5%. The fluctuations are mainly concentrated in Mengshan Dawa area in the northwest of the study area, the southern part of Tangcun reservoir, and Jiujianpeng area. The research results can provide some references for the selection of extraction algorithms for geological hazard easy-happening areas and the prevention of geological hazards in Pingyi county in Shandong province.