Landslide is one of these natural hazards which causes a great amount of financial and human damage annually allover the world. Accordingly, identification of areas with landslide threat for implementation of preventive measures in order to confront against the instability of hillsides for reduction of potential threats and related risks is very important. In this research a new method for classification of landslide risk according to geographical analysis and uncertainty modeling is presented which is based on data mining in previous events. In order to do so, adaptive neuro-fuzzy algorithm which is adjusted by means of sensitivity analysis is used in inferential basis of proposed model, which analyze landside risk efficiently. The selected region for this study is available lands in Alborz province. In proposed method factors like altitude, petrology, gradient, gradient direction, distance to fault and rainfall which are some of the most serious causes of hillside's instability had been inserted and their raster maps produced in GIS context and stored in georeference database. In the next step, areas prone to landslide had been identified according to findings of proposed model and finally in addition to model evaluation according to validation outputs, another round of validation is done by field monitoring of hih-risk regions and interpretation of provided 3D models. Results show that the proposed model with root mean square error of 0.819 and correlation factor of 0.934 has a relatively high accuracy in classification of landslide risk. In addition in landslide risk geographical distribution map inside studied region, the area of landslide-prone area is the highest with respect to total area of province which shows high-risk of Alborz province against landslides.