基于Lasso和BP神经网络的组合预测及其应用
2016-03-07喻胜华张静
喻胜华+张静



摘要:在变量选择的基础上,构建基于Lasso方法和BP神经网络的预测模型,并对我国城乡居民的消费支出进行预测,结果显示:基于Lasso方法和BP神经网络的组合预测精度要明显高于BP神经网络、Lasso方法的预测精度;在2014-2020年,我国农村居民消费增长率有所提升,城镇居民消费增长率减缓,城乡居民消费增长率之间的差距呈下降趋势,但是短期内城乡居民消费差距依然难以缓和。
关键词: 消费;Lasso方法;BP神经网络;预测
中图分类号:F069
The Study on Prediction of Residents Consumption Expenditure
based on Lasso and BP Neural Network
YU Sheng-hua, ZHANG Jing
( School of Economics and Trade, Hunan University, Changsha, Hunan 410079 )
Abstract: On the basis of variable selection, created a multivariate prediction model based on the combination of Lasso method and BP neural network ,and prediction of China's urban and rural residents consumption expenditure. The prediction results showed that:the combination of Lasso method and BP neural network prediction accuracy is higher than that of the BP neural network,the Lasso method,the results also showed that in 2014-2020 years, the growth rate of rural residents consumption has improved,the consumption of urban residents increased slowly, the gap between urban and rural consumption rate showed a downward trend,but the gap between urban and rural consumption is still difficult to ease in the short term.
Key words: consumption; Lasso method; BP neural network; prediction
一、引言及文献综述
近年来,欧美经济复苏乏力,影响世界经济的各种不确定因素有增无减。我国进出口总额的增长速度不如预期,且存在进一步下行的压力,表明世界经济放缓已对我国外贸产生较大的影响,净出口对我国经济增长的拉动已显得力不从心,目前,中国经济主要靠投资和消费拉动,但如果投资规模过大,又会引致结构调整不力、产能过剩和重复建设等问题。为应对极其严峻的国内外经济形势,党和政府多次强调要着力扩大内需,特别是居民消费需求。与世界其他一些国家相比,我国居民的最终消费率仍偏低,属于典型的高储蓄、低消费的国家。发达国家消费对经济增长的贡献率已达70%-80%,国内才30%-40%。从1980年至2013年,最终消费对GDP的贡献率和拉动率分别由71.8%、5.6个百分点下降到50%、3.85个百分点。在当前及今后一段时间,扩大国内消费需求已成为拉动经济增长、保持国民经济持续健康协调发展的重要途径和必然选择。……
