[Objective] A high-precision geological hazard susceptibility evaluation model was determined for the three districts (counties) of Lishi, Shilou, and Liulin in Lüliang City, Shanxi Province in Luliang mountaionous area in order to provide auxiliary decision-making support for regional planning in the area.[Methods] Based on a geographic information system, a sample of 525 historical hazard points and 500 randomly selected non-hazard points in the region were used, and 19 influencing factors of geological hazard were selected. Geographic detectors (GD) were used to judge the relative importance of each factor. Correlation tests and filtering index factors were determined on the Jupyter Notebook platform. Based on the information method (IM), a method was proposed to calculate the amount of information provided by disaster points combined with the amount of information provided by non-disaster points to obtain the improved information method (IIM), and to calculate the weight with the help of the spatial heterogeneity q value of geographic detectors. Six evaluation systems (GD-IIM, GD-IM, GD-CF, IM, CF, and IIM) were established using the certainty factor (CF). The natural breakpoint classification method was used to divide the susceptibility into five, four, and three levels, and the accuracy of the partition results was verified by the seed cell area index (SCAI). The accuracy of the model results was compared with the ROC curve.[Results] After SCAI testing, the models were divided into four levels (very low, low, high, and very high) that met the rationality requirements. The success rate and prediction rate of disaster susceptibility evaluation by the GD-IIM model reached 90.5 % and 85.5 %, respectively. The IIM model exhibited 2 %~4 % greater accuracy than the traditional IM and CF statistical methods.[Conclusion] The bivariate statistical method coupled with geographic detectors produced more accurate results in constructing the vulnerability evaluation prediction model in the study area. Model construction that considered the non-disaster point information was more accurate than the IM model that considered only the disaster point information model. The improved model was suitable for local model construction.