摘要
[目的/意义]随着机器学习算法研究的不断突破,在GDP预测领域中的应用也日益广泛。系统梳理该领域的相关议题,有助于学术研究的纵深发展。[方法/过程]运用文献分析法对公开发表的学术成果进行归纳,梳理机器学习在GDP预测领域研究的进展与脉络。[结果/结论]现有研究逐渐将原本以灰色预测模型、因子模型、传统时间序列和动态因子模型为主的短期预测功能扩展为以神经网络为核心,支持向量机、贝叶斯算法为补充的长期预测功能。同时,各种研究通过对原有参数进行优化、将模型进行对比、利用不同模型进行组合预测等方式来提高预测精度。机器学习与GDP预测的结合,也使建模思想从线性转向非线性,从关注模型参数优化转向与其他方法相结合。除此之外,算法改进与模型运用的多样性、不同模型之间的多维度比较,仍然是该领域的重点和难点。未来研究可以尝试在深入对比各模型的基础上实现更加有效的组合,或挖掘仍未被使用的模型的潜在预测能力。
[Purpose/significance]With the continuous breakthrough in the research of machine learning algo rithm,its application in the field of GDP prediction is becoming more and more extensive.Systematically combing the relevant topics in this field will contribute to the in-depth development of academic research.[Method/process]Using the method of literature analysis,this paper summarizes the published academic achievements,and combs the research progress and context of machine learning in the field of GDP prediction.[Result/conclusion]The short-term prediction function based on grey prediction model,factor model,traditional time series and dynamic factor model is gradually extended to the long-term prediction function with neural network as the core while support vector machine and Bayesian algorithm as the supplement.At the same time,various researches improve the prediction accuracy by comparing the models and using different models to make combined prediction.This also promotes the modeling idea to change from linear to nonlinear,and from focusing on model parameter optimization to combining with other methods.In addition,the diversity of algorithm improvement and model application,and the multi-dimensional comparison between different models are still the focus and difficulty of this field.Future research can try to achieve more effective combination on the basis of in-depth comparison of various models,and design the potential prediction ability of unused models.
作者
马静雯
李树青
夏梦瑶
MA Jingwen;LI Shuqing;XIA Mengyao(School of accounting,Nanjing University of Finance&Economics,Nanjing 210023;College of Information Engineering,Nanjing University of Finance&Economics,Nanjing 210023)
出处
《科技情报研究》
2022年第3期73-94,共22页
Scientific Information Research
基金
国家社会科学基金项目“学术虚拟社区知识交流效率研究”(编号:17BTQ028)。
关键词
机器学习
GDP预测
神经网络
综述
对比分析
machine learning
GDP forecast
neural network
review
comparative analysis
