Landslides are one of the most dangerous natural phenomena. The occurrence of this phenomenon at low speeds and high rates causes financial and human losses without warning signs. Therefore, it is essential to study the geological and anthropogenic factors affecting the occurrence of this phenomenon and determine the potential landslide zones. This study aims to use a supervised convolutional artificial neural network to model landslide potential. For this, evidence maps of seven effective factors in landslide occurrence, including slope, slope direction, geology, precipitation, distance from the fault, height, and density of waterway, were prepared. Then the values in the maps were assigned by continuous fuzzy weights through a logistic function, without data classification to feed the convolution artificial neural network algorithm. For training the network and testing the results, 70% and 30% of training sites, in Oshvand basin, Hamedan province, Iran were used to generate landslide potential model. A prediction-area plot was used to evaluate and quantify the effectiveness of the models produced. The results showed that 70% of the landslides occurred in 30% of the area.
