Subjective Well-Being Prediction Using Data Mining Techniques: Evidence from Chinese General Social Survey

Abstract

Data mining techniques have attracted increasing attentions recently and played more and more important roles in various domains. However, few studies have used these prevalent techniques to explore the rules of subjective well-being for individuals. In this study, a prevalent data mining method, XGBoost, is applied to predict the subjective well-being according to various predictive factors. Feature selection step is implemented to further improve the prediction results and reduce the computational complex based on the importance calculated by XGBoost. An authoritative academic database, Chinese General Social Survey, is used for providing an evidence for classification prediction performance. Moreover, five benchmark models, i.e., logistic regression, support vector machine, decision tree, random forest, and gradient boosting decision tree, are used for comparative analysis based on three evaluation metrics, Accuracy, AUC and F-score. The experimental results indicate that XGBoost outperforms other benchmark models, and feature selection step can improve the prediction performance and reduce the computational time to some extent. In reality, using data mining methods can deeply explore the rule of subjective well-being based on various predictive features, and provide an overwhelming support for improving subjective well-being. Therefore, the methods used in this study are effective and the results provide a support for making society more harmonious.