Theme Evolution Analysis of Public Security Events Based on Hierarchical Dirichlet Process

摘要

In order to analyze the topic evolution in the field of public safety events, this paper proposed a model named Dynamic Micro-blog Hierarchical Dirichlet Process (MD-HDP). First, taking the specific security incidents as the key words, the related texts in micro-blog are extracted precisely by using time features and topic labels. Then, the event texts in each time window are modeled separately to get the corresponding topics for each incident. Finally, the evolution of the subject matter is analyzed of the security incident. The comparison experiments show that the MD-HDP model has a good performance on mining the evolution of the security incident topic and provides more accurate results on the evolution of the security incident topic over time.