Prediction and unsupervised clustering of fertility intention among migrant workers based on machine learning: a cross-sectional survey from Henan, China.

Background: Although China has implemented multiple policies to encourage childbirth, the results have been underwhelming. Migrant workers account for a considerable proportion of China's population, most of whom are of childbearing age. However, few articles focus on their fertility intentions.

Method: From August 3 to August 29, 2023, we conducted a cross-sectional survey in Henan Province, China, which included 18,806 participants. Machine learning was used to construct a predictive model for the fertility intention of migrant workers, and unsupervised clustering was used to explore subgroup classification.

Result: Out of 18,806 participants, only 1057 had fertility intention. We constructed a predictive model for fertility intention based on XGBoost, with an AUC of 0.83. Age, number of children, and marital status are the most important characteristics that affect the fertility intention of migrant workers. Subsequently, unsupervised clustering was conducted on participants without fertility intentions, and it was found that they could be divided into three categories of population. The first group of people is the youngest and mostly unmarried without pregnancy, the second group has the lowest monthly income and self-perceived economic level, and the third group is the oldest and has the highest proportion of women. Regardless of the group, economic and age factors are the main reasons for participants not having a family plan in the near future.

Conclusion: The fertility intention of the migrant workers is at an extremely low level. Improving childcare-related benefits and family support services to reduce the economic and time costs of childcare is an effective measure to reverse fertility intentions.