A New Approach for Stock Price Analysis and Prediction Based on SSA and SVM

作者：Xiao, Jihong; Zhu, Xuehong; Huang, Chuangxia; Yang, Xiaoguang; Wen, Fenghua; Zhong, Meirui^*

来源：International Journal of Information Technology and Decision Making, 2019, 18(1): 287-310.

DOI：10.1142/S021962201841002X

• Stock price
• singular spectrum analysis
• support vector machine
• combined model

Stock price exhibits distinct features during different time scales due to the effects of complex factors. Analyzing these features can help delineate the mechanisms that determine the stock price and enhance the prediction accuracy of the stock price. By using singular spectrum analysis (SSA), this paper first decomposes the original price series into a trend component, a market fluctuation component and a noise component to analyze the stock price. The economic meanings of the three components are identified as a long-term trend, effects of significant events and short-term fluctuations caused by noise in the market. Then, to take into account the features of the above three components to the stock price prediction, a novel combined model that integrates SSA and support vector machine (SVM) (e.g., SSA-SVM) is proposed. Compared with SVM, adaptive network-based fuzzy inference system (ANFIS), ensemble empirical mode decomposition-ANFIS (EEMD-ANFIS), EEMD-SVM and SSA-ANFIS, SSA-SVM demonstrates the best prediction performance based on four criteria, indicating that the proposed model is a promising approach for stock price prediction.

• 出版日期2019-1
• 单位中国科学院; 中国科学院数学与系统科学研究院; 中南大学; 长沙理工大学