Research on Suicide Identification Method Based on Microblog “Tree Hole” Crowd

Suicide has always been a key issue of social and health organizations and research of scholars. In recent years, with the increasing popularity of Internet and social media, more and more people record their lives and feelings in social media, even more, they publish suicide speeches, especially in the youth. The microblog “tree hole” is generated because a depressed patient has left, a large number of potential depression patients or suicide prone people to leave messages under her account until today, including venting negative emotions, suicide messages and even committing suicide together. In view of this situation, the paper analyzes the actual needs and business scenarios of monitoring and identifying online depression suicide messages. Based on the data of “tree hole” message on microblog, this paper designs and implements a multi-channel convolutional neural network social suicide warning model. By mining the time of message of “tree hole” people in microblog, the feature matrix is obtained after quantifying some knowledge elements; After word orientation quantization, word vector matrix and word vector matrix are obtained. Feature matrix and text vector matrix are used as three inputs of convolutional neural network respectively to identify and warn social suicide. The model is proved to be better by experiments and improve the accuracy of early warning.