Citation

Zhuang T, li D, Tan W, Wang G. Harbin Gongcheng Daxue Xuebao 2019; 40(11): 1890-1895.

Copyright

(Copyright © 2019, Gai da xue)

DOI

10.11990/jheu.201809034

PMID

unavailable

Abstract

To solve the potential suicide tendency problem in social media, taking Sina Weibo as the research object, a hierarchical classification scheme based on domain knowledge is designed in this study, and a classification model of hierarchical support vector machine based on the classification scheme is proposed. This model provides early identification of the group with high suicide risk and can be used for suicide tendency detection and intervention to reduce suicide probability. By optimizing the parameters of the model layer by layer, the impact of unbalanced data on the experimental result is reduced. By comprehensively considering the emotional state of the users, the emotional dictionary is continuously expanded. The experimental results show that the prediction accuracy of the model for suicide probability reaches 0.848, which confirms the model can effectively predict the suicidal tendency of Weibo users. Meanwhile, it was found that a relationship exists between the post time of microblog and the suicide risk probability, and the relationship can be fitted by a normal curve. © 2019, Editorial Department of Journal of HEU. All right reserved.

Language: zh

Keywords

Forecasting; Learning systems; Social networking (online); Support vector machines; Classification (of information); Sentiment analysis; Machine learning; Microblog; Classification models; Classification scheme; Hierarchical classification; Hierarchical support vector machines; Micro-blog; Predicting and analyzing; Prediction accuracy; Suicide tendency prediction; Support vector machine(SVM); Tendency prediction