基于深度学习的大学生幸福感测度分析  

Analysis of College Students’ Happiness Perception Based on Deep Learning

作　　者：付瑞鑫 张莉[1] 屈启兴[1] 

机构地区：[1]对外经济贸易大学信息学院,北京

出　　处：《建模与仿真》2022年第6期1424-1433,共10页Modeling and Simulation

摘　　要：大学生情绪容易受到各类事件影响,学业求职的激烈竞争和疫情的突如其来导致其幸福感明显降低,利用社交媒体分析大学生幸福感引起了国内外学者的关注。本文提出一种利用社交媒体数据,基于深度学习ALBERT-TextCNN模型的大学生幸福感测度分析方法。首先使用ALBERT预训练语言模型将社交媒体文本描述转化成向量表示,提取文本描述中的关键特征,然后将提取到的特征送入TextCNN模型进行分类预测,得出社交媒体文本的情感极性,最后将积极情感文本占比作为大学生用户的幸福感指数。在公开微博文本数据集上进行实验,ALBERT-TextCNN模型在情感极性分类预测上准确率、精确率、召回率和F1值均达到较高水平,同时训练时间短成本低。最终本文利用此模型确定了北京某高校466名大学生的幸福感指数情况。College students’ emotions are easily affected by various events. The fierce competition for academic job hunting and the sudden outbreak of the epidemic have led to a significant decrease in their happiness. The use of social media to analyze college students’ happiness has attracted the attention of scholars at home and abroad. This paper proposes an analysis method of college students’ happiness perception based on deep learning ALBERT-TextCNN model using social media data. First, the ALBERT pre-trained language model is used to convert the social media text description into a vector representation, and the key features in the text description are extracted, and then the extracted features are sent to the TextCNN model for classification and prediction, and the emotional polarity of the social media text is obtained. Finally, the proportion of positive emotional texts is taken as the happiness index of college students. Experiments were carried out on the public microblog text dataset, and the ALBERT-TextCNN model achieved a high level of accuracy, precision, recall, and F1 value in sentiment polarity classification prediction, and at the same time, the training time was short and the cost was low. Finally, this paper uses this model to determine the happiness index of 466 college students in a university in Beijing.

关 键 词：幸福感测度 ALBERT模型 TextCNN模型 社交媒体 

分 类 号：G64[文化科学—高等教育学]