Crime Detection Using Latent Semantic Analysis and Hierarchical Structure

摘要

We make efforts to help the investigator discover the hidden conspirators. In the criminal cases, the investigators or the police have to make full use of the messages or spoken documents data that they record in files. Thus, mining the latent information from messages is vital to them. In Information Retrieval area, Latent Semantic Analysis (LSA) is an important method for query matching which can discover the underlying semantic relation or similarity between words and topics. We introduce a network hierarchical structure to analyze the original message network, making the analysis conveniently as well as ensuring the connectivity of the inner network connection of all the conspirators. For this purpose, we use LSA to measure the similarities between topics and Crime Prototype Vector, and the similarities will be used as the weights of the paths in the network hierarchies and calculate the suspicious degrees.