社会与工程心理学研究室知识库

Suicide is one of the increasingly serious public health problems in modern society. Traditional suicidal ideation detection using questionnaires or patients’ self-report about their feelings and experiences is normally considered insufficient, passive, and untimely. With the advancement of Internet technology, social networking platforms are becoming increasingly popular. In this paper, we propose a suicidal ideation detection method based on multi-feature weighted fusion. We extracted linguistic features set that related to suicide by three different dictionaries, which are data-driven dictionary, Chinese suicide dictionary, and Language Inquiry and Word Count (LIWC). Two machine learning algorithms are utilized to build weak classification model with these three feature sets separately to generate six detection results. And after logistic regression, to get the final weighted results. In such a scheme, the results of model evaluation reveal that the proposed detection method achieves significantly better performance than that use existing feature selection methods.