In this work, it has been verified the ability of a computational tool known as artificial neural networks to predict landslides triggered by rains in mountainous areas. The neuronal model was trained and verified for a tropical mountainous location in the city of Medellin, Colombia. The construction of the database, composed by 2054 samples, consisted in obtaining nine representative slopes and twenty typical rains of the study site, characterized by their duration and return period. There have been combined data of nine slopes and twenty rainfalls, there by obtaining the fluid pressure profile in a finite element code, in order to get the safety in each of the analyzed cases. To obtain shear strength values for the simulations, the parabolic envelope proposed by Lade (2010) was used. Thus, the input data to the neural network are the slope and the precipitation and the output is the safety factor. It has been verified the ability of the neural networks to learn specific information of the studied problem and give an answer, with minimum error, for any other condition, generalizing the problem and allowing the application of the tool in a location with similar conditions.