摘要
现有的区域经济预测模型指标变量存在冗余,且忽略了行业、区域等静态变量对预测结果的影响,导致预测效率不高.针对上述问题,提出了基于改进Transformer时序算法的区域经济预测模型.首先利用Copula函数对传统Transformer模型进行优化(Coformer);其次选取区域经济的影响指标,对其进行主成分分析,去除冗余信息;然后将降维后的指标变量和静态变量作为Coformer的输入,对变量进行编码,并通过多头注意力机制增强重要信息,最后用解码器对编码的变量解码,利用Softmax输出区域历年生产总值序列的预测结果.实验结果表明,所提模型的预测准确率为0.908,比另外三种模型分别提高了15.9%、12.3%和6.7%,表现出了优异的预测性能.
The existing regional economic forecasting model indicators are redundant and ignore the impact of static variables such as industry and region on the forecast results,leading to low forecasting efficiency.In response to the above issues,a regional economic forecasting model based on an improved Transformer time series algorithm is suggested.Firstly,the traditional Transformer model is optimized using a Copula function(Coformer);secondly,the impact indicators of the regional economy are selected,and principal component analysis is performed on them to remove redundant information;then,the reduced-dimensional indicator variables and static variables are used as the input of the Coformer,and the variables are encoded.Finally,the decoder decodes the encoded variables and uses Softmax to output the prediction results of the regional gross domestic product series over the years.The experimental outcome indicates that the designed model has an accuracy of 0.908,which is 15.9%,12.3% and 6.7% higher than the other three models,respectively,demonstrating excellent predictive performance.
作者
刘海宏
刘敏
朱岸青
Liu Haihong;Liu Min;Zhu Anqing(School of Economics and Management,Guangzhou Nanyang Polytechnic College,Guangzhou 510900,China;School of Computer and Software,Southwest Petroleum University,Nanchong 637001,China;School of Management,Jinan University,Guangzhou 510000,China)
出处
《南京师大学报（自然科学版）》
CAS
北大核心
2024年第4期118-125,共8页
Journal of Nanjing Normal University(Natural Science Edition)
基金
广东省哲学社会科学规划项目(GD22XYJ28)
广东省教育厅项目(2021GXJK595、2021TSZK021)。
作者简介
通讯作者:刘海宏,博士,副教授,研究方向:人工智能与大数据,经济管理.E-mail:8928114@qq.com。
