[Objective] The influence of different connection methods on Benggang susceptibility was studied in order to provide a theoretical basis for the refined spatial prediction of regional Benggang in order to provide scientific supports for reasonable, economical, and efficient disaster prevention and reduction. [Methods] A random forest (RF) model was used to select two different connection methods: the improved frequency ratio (ALSA) and the traditional frequency ratio (FR). Annual average rainfall, annual average rainfall erosivity, soil erodibility, slope length factor, clay content, slope, sand content, and standardized VH channel backscatter coefficient (representing the nature of the surface cover) were used as evaluation indexes of Benggang susceptibility in Yongfeng Township, Xingguo County of Jiangxi Province. [Results] ① The accuracy of the ALSA-RF model was greater than the accuracy of the FR-RF model, and the values of ACC and AUC were 83.89% and 0.893 0, respectively; ② The extremely high and high susceptibility areas in Yongfeng Township were mainly located in the southwest, and this result was in good agreement with the actual situation. [Conclusion] The improved frequency ratio connection method better reflected the distribution law of regional Benggang susceptibility than the traditional frequency ratio method, and provided a reference for similar problems in the field of landslide evaluation.