基于面板计量与机器学习模型的居民消费水平影响因素的分析与预测

An Analysis and a Prediction of Influencing Factors of Residents’Consumption Level Based on Panel Measurement and a Machine Learning Model

居民消费水平是体现一个国家居民生活质量和国家发展能力的重要指标.本文对我国31个省份2006—2020年的居民消费支出进行了随机效应的回归分析,研究了居民消费水平和居民可支配收入的关系;以1978—2021年我国居民消费支出数据为基础,构建Stacking机器学习模型进行训练,对不同滑窗和是否使用交叉验证进行了对比实验;使用精度最高的模型对未来5年(2022—2026年)居民消费支出进行了预测,实验结果表明:1)提高居民可支配收入、稳定物价和经济的高质量发展有利于居民消费水平的提高;2)在对居民消费水平预测中,使用交叉验证方法和以4年为滞后特征的预测模型更加精准,未来5年居民消费水平将缓慢持续上升.

The level of residents’consumption is an important index that reflects the quality of their life and the capacity of national development.In this paper,a random effect regression analysis was carried out on the consumption expenditure of 31 provinces during the 2006-2020 periods,and the relationship between consumption level and disposable income was studied;based on the consumer expenditure data of Chinese residents from 1978 to 2021,a Stacking machine learning model was constructed to train it.The comparative experiments of different sliding window sizes and whether cross verification should be used were explored.The most accurate model to forecast household consumption expenditure in the next 5 years(2022-2026)was used.The experimental results show that:1)the increase of disposable income,stable prices and high-quality economic development are conducive to the improvement of residents’consumption level;2)when predicting residents’consumption level,the cross-verification method and the prediction model with 4-year lag features are more accurate,and residents’consumption level will rise slowly and continuously in the next five years.