Modeling a road accident hazard zoning map to identify high-risk areas is a very effective step to reduce the resulting casualties. Due to the dynamic nature of many of the factors affecting the identification of these areas, traditional zoning mapping does not seem to be effective. In the field of ubiquitous modeling in the framework of the GIS, it is possible to produce a separate map at any time, any place, for any user and under any circumstances that is more compatible with his or her changing individual and environmental conditions. In this study, in a hybrid model of data mining methods based on 22 environmental and individual contexts the probability of accident risk was calculated for each user. Thus, by obtaining the accident data recorded in Tabriz Marand road in 2019, the required pre-processing was done based on T2 statistic and PCA method. Then, based on GRNN and the collected data, the optimal network was trained and then evaluated with test data. In addition, eight different scenarios were designed for this case study and at 3008 points of this road, the risk of accident was predicted for each scenario. The results showed more than 90 % accuracy for the proposed model of this study. According to the results of different scenarios, the 5 km area around customs and 3 km around Sufian city have the highest risk in this road, which can be verified by 50 accidents happened in this area during the mentioned period.