Influence propagation in social networks has recently received large interest. In fact, the understanding of how influence propagates among subjects in a social network opens the way to a growing number of applications. Many efforts have been made to quantitatively measure the influence probability between pairs of subjects. Existing approaches have two main drawbacks: (i) they assume that the influence probabilities are independent of each other, and (ii) they do not consider the actions not performed by the subject (but performed by her/his friends) to learn these probabilities. In this paper, we propose to address these limitations by employing a deep learning approach. We introduce a Deep Neural Network (DNN) framework that has the capability for both modeling social influence and for predicting human behavior. To empirically validate the proposed framework, we conduct experiments on a real-life (offline) dataset of an Event-Based Social Network (EBSN). Results indicate that our approach outperforms existing solutions, by efficiently resolving the limitations previously described.