Embodiments of the invention provide a data processing method and device. The method comprises the following steps of: obtaining to-be-processed social behavior data flow; pre-processing the social behavior data flow and converting the social behavior data flow from a data space to low-dimensional feature vectors; and inputting the low-dimensional feature space vectors to a plurality of layers of restricted Boltzmann machine RBM stacks to carry out calculation processing, so as to complete the extraction of concealed features in the social behavior data flow. By adopting the data processing method and device, the abstract concealed features in the social behavior data flow can be automatically extracted through the plurality of layers of RMB stacks, so that the efficiency is improved and the research and development costs are decreased.