摘要
金融时间序列数据具有非线性、非平稳、高噪声等复杂特征,且随着移动互联网、人工智能的快速发展,海量结构化与非结构化数据不断产生,数据间的关联模式日益复杂。在此背景之下,构建科学合理的金融时间序列数据预测模型,充分挖掘金融时间序列数据隐含的重要信息至关重要。为此,在梳理金融时间序列数据预测的计量经济学方法、机器学习算法的基础上,着重分析深度学习应用于金融时间序列数据预测的理论基础与实证应用的相关文献,以期为大数据与人工智能背景下的金融时间序列数据预测以及多学科交叉融合研究提供相关借鉴。
Financial time series data has complex characteristics such as non-linear,non-stationary and high noise.With the rapid development of mobile internet and artificial intelligence,massive structured and unstructured data are constantly generated,and the relation among them is increasingly complex.So,it is very important to build a scientific and reasonable financial time series data prediction model and fully mine the important information hidden in the financial time series data.Compared with econometric model and machine learning,deep learning algorithm,which has achieved great success in image recognition,self-driving,natural language and many other fields of artificial intelligence,is more suitable for financial time series data prediction under the background of big data and artificial intelligence.
作者
闫洪举
YAN Hongju(Postdoctoral Research Station of Agricultural Bank of China,Beijing 100005,China;School of Economics,Peking University,Beijing 100871,China)
出处
《金融教育研究》
2021年第3期33-41,共9页
Research of Finance and Education
关键词
时间序列数据
计量经济学
机器学习
深度学习
Time series data
Econometric model
Machine learning
Artificial intelligence