<p>Due to the variable and complex behavior labels of power market players, the multi-dimensional portrait prediction of power market players’ behavior is not effective. In order to solve this problem, this paper proposes a multi-dimensional profiling prediction method of power market subject behavior based on deep neural network. The distributed web crawler architecture is used to capture the behavioral data of power market entities, and the K-Means algorithm is adopted to cluster the captured data, so as to realize the behavioral mining of power market entities. According to the results of data mining, Text Rank is used to generate the main body label of electricity market, and the label is optimized. The cyclic neural network in the deep neural network is used to extract the multi-dimensional characteristics of power market players’ label attributes, and the multi-dimensional portrait prediction of power market players’ behavior is realized by combining the gradient lifting decision tree. The experimental results show that this method can realize the fast and accurate prediction of multi-dimensional behaviors of power market players.</p>