Stacking Model for Optimizing Subjective Well-Being Predictions Based on the CGSS Database

Abstract

Subjective Well-Being (SWB) is an important indicator reflecting the satisfaction of residents’ lives and social welfare. As a prevalent technique, machine learning is playing a more significant role in various domains. However, few studies have used machine learning techniques to study SWB. This paper puts forward a stacking model based on ANN, XGBoost, LR, CatBoost, and LightGBM to predict the SWB of Chinese residents, using the Chinese General Social Survey (CGSS) datasets from 2011, 2013, 2015, and 2017. Furthermore, the feature importance index of tree models is used to reveal the changes in the important factors affecting SWB. The results show that the stacking model proposed in this paper is superior to traditional models such as LR or other single machine learning models. The results also show some common features that have contributed to SWB in different years. The methods used in this study are effective and the results provide support for making society more harmonious.