基于LSTM神经网络的人口预测研究

Research on Population Prediction Based on LSTM Neural Network

摘要: 本文对人口数据进行整理，探究全国总人口数与相关影响因数之间的分布关系。我国人口数量的影响因数有自然资源、社会环境、其他因素(男女占比、人口老龄化问题、相关政策等)。本文提出LSTM神经网络预测模型，以我国人口统计数据为研究样本，整合多项关键指标，构建LSTM预测模型，并将该模型与传统灰色预测GM(1, 1)模型、阻滞增长模型开展对比实验。实验结果表明，所构建的LSTM神经网络模型预测精度更高，其结果通过预测2020年我国总人口数据与已知2020年我国总人口数据进行对比得到，同时预测2030年我国总人口数据，该研究成果为人口动态预测提供了可靠的技术方法。

Abstract: This paper organizes population data to investigate the distributional relationships between the total national population and its relevant influencing factors. The factors affecting China’s population size include natural resources, the social environment, and other elements (such as the gender ratio, population aging, and relevant policies). This paper proposes an LSTM neural network-based prediction model, utilizing China’s population statistics as the research sample. By integrating multiple key indicators, an LSTM prediction model is constructed, and comparative experiments are conducted between this model and traditional forecasting approaches, namely the Grey Prediction GM(1, 1) model and the Logistic Growth Model. The experimental results indicate that the constructed LSTM neural network model achieves higher prediction accuracy. This conclusion is validated by comparing the model’s predicted total population data for China in 2020 with the known actual data for the same year. Furthermore, the model is used to forecast China’s total population for 2030. The findings of this study provide a reliable technical approach for dynamic population prediction.