基于梯度因子的ARMA-GARCH股票价格预测模型研究  

ARMA- GARCH Model Based on Gradient Factor for Stock Price Forecast

作　　者：张贵生[1,2] 张信东[1,2] 

机构地区：[1]山西大学管理与决策研究所,山西太原030006 [2]山西大学经济与管理学院,山西太原030006

出　　处：《山西大学学报（哲学社会科学版）》2016年第1期115-122,共8页Journal of Shanxi University(Philosophy and Social Science Edition)

基　　金：国家自然科学基金面上项目(71371113);教育部人文社会科学研究项目(13YJA790154)

摘　　要：股票价格时间序列的不确定性是股票市场具有复杂运动规律的综合外在表现形式,通过科学建模对其展开预测性研究已经成为金融研究领域的一大热点。作为一种经典的金融时序回归方法,ARMA-GARCH模型在处理股指数据时却并未融合股票价格行为背后所蕴含的经验知识,在一定程度上影响了预测模型泛化能力的进一步提高。本文把时序历史数据变化趋势的微分信号和股票市场的隔夜跳空开盘信息综合提取为一个梯度因子并引入到传统ARMA-GARCH模型中,建立了基于梯度因子的G-ARMA-GARCH模型。通过对上证综指、深证成指、香港恒生和标准普尔500指数收盘价收益率的实证研究表明,融合了梯度因子信息的G-ARMA-GARCH模型预测精度显著提高,而且具有良好的稳定性。The uncertainty of the stock price time series is the comprehensive external manifestation of the stock market with high complexity. And the forecast research for the stock price by scientific modeling has been a hotspot in the field of financial research area. However,as a kind of classic financial time series regression method,the ARMA- GARCH model dose not incorporate the experience knowledge hidden in the stock price series data in dealing with the stock price series data,which influences the model＇s generalization ability to some extent. In this paper,by integrating the differential signal of the historical data＇s development tendency in time series and the information of overnight gap opening in stock market,the author extracts a gradient factor,which is introduced to the traditional ARMA- GARCH model to build a new one called G- ARMA- GARCH. Experimental results based on the Shanghai composite index,Shenzhen component index,Hong Kong Hang Seng and Standard Poor＇s 500 closing index show that the G- ARMA- GARCH model combined with the gradient factor information improves significantly in its prediction accuracy,what＇s more,it is of promising stability.

关 键 词：股票价格预测 ARMA-GARCH 梯度因子 

分 类 号：F830.91[经济管理—金融学]