EPAK: A Computational Intelligence Model for 2-level Prediction of Stock Indices

International journal of computers, communications and control

Ping He Pan¹  Li Tang²  Yong Yi Yao³ 

[1] Intelligent Finance Research CenterChongqing Institute of Finance;School of Management and Economics, University of Electronic Science and Technology of ChinaTianfu college of Southwestern University of Finance and Economics;Tianfu college of Southwestern University of Finance and Economics

关键词: empirical mode decomposition;    principal component analysis;    affinity propagation;    k-nearest neighbor;    time series;    stock index prediction;   

DOI  :  10.15837/ijccc.2018.2.3187

学科分类：计算机科学（综合）

来源: Universitatea Agora

【 摘 要 】

This paper proposes a new computational intelligence model for predicting univariate time series, called EPAK, and a complex prediction model for stock market index synthesizing all the sector index predictions using EPAK as a kernel. The EPAK model uses a complex nonlinear feature extraction procedure integrating a forward rolling Empirical Mode Decomposition (EMD) for financial time series signal analysis and Principal Component Analysis (PCA) for dimension reduction to generate information-rich features as input to a new two-layer K-Nearest Neighbor (KNN) with Affinity Propagation (AP) clustering for prediction via regression. The EPAK model is then used as a kernel for predicting each of all the sector indices of the stock market. The sector indices predictions are then synthesized via weighted average to generate the prediction of the stock market index, yielding a complex prediction model for the stock market index. The EPAK model and the complex prediction model for stock index are tested on real historical financial time series in Chinese stock index including CSI 300 and ten sector indices, with results confirming the effectiveness of the proposed models.

【 授权许可】

Free