The Determinants of Hukou Acquisition Intentions among Migrants in Tianjin: A Study Using XGBoost and SHAP

In this paper, the author investigates the key factors influencing migrants’ intentions to get hukou in Tianjin according to the 2017 China Migrants Dynamic Survey (CMDS), which includes 5,000 questionnaires, the study uses XGBoost to find significant determinants. Uses SHAP (SHapley Additive exPlanations) to interpret the relative importance. The study researches many factors, including characteristics of the migration destination, individual attributes, factors related to the hukou origin city, and stay wish. The machine learning compared to traditional logistic regression methods, XGBoost can capture more complex nonlinear relationships between different periods in one feature, and supply a more precise analysis of how different factors and how the feature change influence Hukou Acquisition Intentions across different levels. Additionally, unlike traditional studies on hukou acquisition intentions, machine learning gives more deeper insights into the different self characteristics of willing to obtain hukou in Tianjin.