The Classification of Chinese Personal Income Level Based on Bayesian Network

In recent years, great changes have taken place in economy and society in China. However, income inequality is becoming more serious and it needs to be paid more attention. Therefore, the analysis of factors that affect income is important. The Bayesian network is a common method to study causal relationships among different variables. The paper analyzed personal annual income in 2016 in China based on the data of Chinese General Social Survey (CGSS) with the Bayesian network (BN). The research is to study the relationships among 14 income related factors and classify the personal income level. Based on the per capita disposable income in 2016 in China (23821 yuan), personal income was divided into two categories: High Income (personal income was greater than 23821) and Low Income (personal income was smaller than 23821). Then we applied BN to classify the level of personal income. The predicted classification results with Bayesian network were compared with those with Naïve Bayesian method. It could be found that BN could not only reflect the causal relationships among 14 variables, but also have higher prediction accuracy in this income problem.