Abstract:In recent years, the overall birth rate of the whole country has been declining, but the rate of reduction of birth rate varies among cities at different levels. It is urgent to set an appropriate threshold for dividing the level of birth rate decline in each region, and to analyze the influencing factors of birth rate changes. This paper selects the indicator data of 287 prefecture-level cities from 2018 to 2019, combines various dimensionality reduction methods to model on the basis of logistic regression, and divides the birth rate drop into three categories: less than 0, less than 5%, and less than 10%. a threshold. The results show that the variables retained by the optimal models under different thresholds are concentrated on indicators such as ecology, economic level, education and culture, and when the threshold is 10%, the model has the strongest classification ability and the best indicators.

Key Words: birth rate change;threshold;influencing factors;dimensionality reduction