【摘要】 为实现大学生心理危机的有效预警,以新浪微博平台为数据源,构建了包含1000位用户及其114790条帖子的数据集,并提出一种用户级融合多维度特征的社交媒体心理预警模型(U-MFSPM)。该模型通过文本语义模块[双向门控循环单元(Bi-GRU)与自注意力机制]提取用户级文档语义特征,情感特征模块[情感词典结合双向长短期记忆(Bi-LSTM)与注意力机制]生成用户多维情感向量,网络行为模块提取性别、发帖时间分布、粉丝数等特征,融合后输入多层感知机(MLP)完成心理状态分类。实验结果表明,该模型的F1-score达92.7%,优于传统模型。研究表明,基于社交媒体数据整合多维度特征可为大学生心理危机预警提供新的有效思路。更多还原

【Abstract】 To achieve effective early warning of college students’ psychological crises, this study constructed a dataset containing 1000 users and their 114 790 posts using Sina Weibo as the data source, and proposed a User-level Social media Psychological warning Model integrating Multi-dimensional Features(U-MFSPM). The model extracts user-level document semantic features through a text semantic module(Bi-GRU(Bidirectional Gated Recurrent Unit)combined with a self-attention mechanism), generates user multi-dimensional emotional vectors via an emotional feature module( emotional dictionary combined with Bi-LSTM( Bidirectional Long Short-Term Memory) and an attention mechanism), and extracts features such as gender, post time distribution, and the number of followers through a network behavior module. After feature fusion, the integrated features are input into a Multi-Layer Perceptron(MLP)to complete the classification of psychological states. Experimental results show that the F1-score of this model reaches92.7%, which is better than that of traditional models. The study indicates that integrating Multi-Dimensional Features based on social media data can provide a new and effective approach for early warning of college students ’psychological crises.更多还原