Spatiotemporal analysis of Population aging based on Geographically and temporally neural network weighted regression
Download as PDF
DOI: 10.25236/etmhs.2024.076
Author(s)
Zilin Shen, Xin Fu, Fan Wang, Duodan Chen
Corresponding Author
Zilin Shen
Abstract
Since the 21st century, the entire world has entered an era of population aging, with a marked deepening of the aging process in developed countries. Therefore, taking into account the two dimensions of time and space to study aging has gradually become a hot issue. However, to analyze the aging rate and its influencing factors from the perspective of regression analysis, we will encounter the problem of how to solve the spatiotemporal non-stationarity. The GWR and GTWR models might encounter challenges when addressing the intricate nonlinear interaction between time and space in reality. Based on this, this paper uses the newly proposed Geographically and temporally neural network weighted regression (GTNNWR) model. By constructing a fully connected neural network model with multiple hidden layers and combining with Dropout technology and other techniques, GTNNWR model can effectively generate space-time weight matrix, and then solve the spatiotemporal non-stationarity. This paper consults the statistical yearbook of the National Bureau of Statistics of China from 2000 to 2020 and used VIF test to select 7 variables without multicollinearity. GTNNWR model is used to analyze aging and its influencing factors, and the results are compared with OLS, GWR and GTWR models. It is proved that GTNNWR has better performance than other models in modelling the spatiotemporal non-stationarity of population aging, and the influencing factors of population aging are analysed according to the results of the final model.
Keywords
population aging; spatiotemporal non-stationarity; GTNNWR