The invention discloses a judgment document analysis method based on a criminal behavior chain, and the method comprises the following steps: step 1, extracting information from the fact description of a case, declaring the concept and construction rule of the criminal behavior chain, and constructing the criminal behavior chain; step 2, obtaining semantic information from fact description of a case by using a TextCNN; step 3, based on the criminal behavior chain, performing vector representation on the criminal behavior chain by using a Bi-LSTM method, and performing text classification; andstep 4, proposing a judgment document analysis method of a neural network structure for splicing and outputting a result of the TextCNN method and a result of the Bi-LSTM method, wherein an identification result can be used for supporting various auxiliary judgment applications. According to the method disclosed by the invention, the associated information of the criminal behavior chain is fully utilized, the defects of a traditional legal judgment document analysis method are avoided, the judgment document analysis efficiency is improved, and a technical support is provided for solving modelsand methods for assisting judgment such as legal provision recommendation and class case pushing.