摘要
目的:基于二维深度学习算法提出大学生负性情绪预警模型,增强大学生负性情绪识别的准确性和稳定性。方法:基于13所高校56837名大学生3类15项特征数据,构建“时间-测度”基于特征灰度的二维卷积神经网络模型(FG-CNN),比较该模型与6类机器学习模型在负性情绪同期预警和跨期预警上的效果。结果:FG-CNN在同期预测上具有94.93%的准确率和0.976的AUC值,在跨期预测上具有91.18%的准确率和0.939的AUC值,均优于一维机器学习结果。结论:学生个体属性和行为模式具有负性情绪预警信息含量,二维深度学习算法可以提高预警精度,为大学生负性情绪认知提供了大数据支撑。
Objective To develop an early warning model of college students'negative emotions based on twodimensional deep learning algorithm which enhances the recognition accuracy and stability of college students'negative emotions.Methods A"Time-Measure"Feature Grayscale-based two-dimensional Convolutional Neural Network(FG-CNN)model was constructed based on 3 types and 15 items of data from 56,837 college students in 13 universities to compare the effectiveness of this model with 6 types of machine learning models for simultaneous and intertemporal warning of negative emotions.Results FG-CNN had 94.93%accuracy and 0.976 AUC in the simultaneous prediction,and 91.18%accuracy and 0.939 AUC in the simultaneous prediction,both of which were better than the one-dimensional machine learning results.Conclusion Students'individual attributes and behavioral patterns have information for negative emotion early warning.Two-dimensional deep learning algorithm can improve the early warning accuracy and provide big data support for college students'negative emotion perception.
作者
程欣炜
薛松
CHENG Xinwei;XUE Song(School of Economics,Nanjing University of Posts and Telecommunications,Nanjing 210023,Jiangsu Province,China;School of Psychology,Nanjing Normal University)
出处
《中国数字医学》
2023年第8期51-56,共6页
China Digital Medicine
基金
国家自然科学基金青年项目-负性情绪影响冲突控制的神经机制(31800915)。
关键词
负性情绪
预警模型
卷积神经网络
特征灰度
机器学习
Negative emotion
Early warning model
Convolutional neural network
Feature grayscale
Machine learning
