基于二重SVM模型的移动互联网用户情绪预测方法

A Mobile Internet User Emotion Prediction Method Based on Dual SVM Model

摘要: 针对移动互联网用户情绪预测问题，收集了40名被试观看偏好以及不偏好视频的主观情绪以及实时心率波动的数据。独立样本T检验结果表明，不同情绪下心率变化幅度存在显著差异，验证了心率在鉴别用户情绪的可行性。在此基础上，提出采用机器学习构建面向移动互联网用户的情绪预测模型。所构建的二重SVM模型对三类情绪(开心、中性和难过)的鉴别成功率均在75%以上。研究结果还表明实时心率数据能够较好地反映出移动互联网用户的情绪变化。

Abstract: To solve the problem of mood prediction of mobile Internet users, the subjective mood and real-time heart rate fluctuation data of 40 subjects were collected. Independent sample t-test results show that there are significant differences in the range of heart rate changes under different emotions, which verifies the feasibility of heart rate in identifying users’ emotions. On the basis of this, the paper proposes to use machine learning to build an emotion prediction model for mobile Internet users. The success rate of the three kinds of emotions (happy, neutral and sad) was more than 75%. The results also show that real-time heart rate data can better reflect the emotional changes of mobile Internet users.