[Objective] The present study was performed to develop a geological disaster susceptibility evaluation model for predicting frequent geological disasters in the Taishan area. The aim was to use the results of this susceptibility evaluation as reference for preventing and managing geological disasters in this area. [Methods] Our analyses focused on the Taishan area, employing a method in which the certainty factor (CF) model was coupled with a random forest (RF) model optimized using the particle swarm optimization (PSO) algorithm to evaluate the geological disaster susceptibility in the research area. This method uses the CF model to calculate the sensitivity values of the factors influencing geological disasters, which are then used as attribute values for model training. The PSO algorithm was introduced to optimize the parameters of the RF model, thereby improving the accuracy and precision of the model in predicting geological disasters. Eleven influencing factors, including slope, distance to roads, land-use type, and vegetation index, were selected. The Pearson correlation coefficient method and a multicollinearity check were used to screen and optimize these influencing factors. The precision of the trained model was evaluated using ROC and PR curves. [Results] Compared with those of the single SVR, single RF, and CF-PSO-SVR models, the CF-PSO-RF coupled model significantly improved the proportion of extremely-high-susceptibility areas by 10.55%, 10.04%, and 5.08%, respectively, increased the AUC values by 14%, 5.1%, and 1.7%, respectively, and enhanced the average precision (AP)accuracy by 11.7%, 4.4%, and 1.2%, respectively. The prediction results revealed that the regions with high and extremely high susceptibility to geological disasters were mainly located in the Taishan scenic area, Northern Daiyue District, and other regions with significant topographic relief and steep slopes, covering 28.05% of the area and encompassing 60.1% of the geological disaster points. In contrast, regions with low and very low susceptibility were primarily distributed in flat areas, such as construction and farm lands, accounting for 59.26% of the total area. [Conclusion] The accuracy of the CF-PSO-RF coupled model for disaster susceptibility evaluation was notably higher than that of the single models; further, its precision was superior to that of the CF-PSO-SVR model. These evaluation results are consistent with actual conditions.