基于人工智能的中国股票价格预测与异质性研究  

Research on Chinese Stock Price Prediction and Heterogeneity Based on Artificial Intelligence

作　　者：张琳 ZHANG Lin

机构地区：[1]北京工业大学经济与管理学院 [2]北京大学数量经济与数理金融教育部重点实验室

出　　处：《暨南学报（哲学社会科学版）》2023年第3期123-132,共10页Jinan Journal(Philosophy and Social Sciences)

基　　金：国家社会科学基金重大项目“货币政策分配效应与缩小收入和财富差距的有效路径研究”(20&ZD105);北京市教育委员会社科计划一般项目“北京金融环境下居民资产选择与价格水平影响研究”(SM202110005014)。

摘　　要：本文构建一个基于人工智能的中国股票价格预测模型,并探究预测效果异质性的原因。根据理论及现实,股票价格主要受利率、市场行为、技术指标以及公司价值因素等影响。对沪深两市具有代表性的个股价格进行预测的实证研究发现:股价预测值与真实值多数时间吻合,日度数据模型与月度数据模型预测精度偏差平均值分别落在1.7%和8%的水平内,且引入控制变量利率的稳健性检验并没有明显的变化产生。进一步的异质性研究给出,模型变量——自身股价、成交量、估值,是股票价格预测效果差异化的理论及现实影响因素。从而为我国政策当局、市场投资指数以及金融机构提供参考、建议和依据,共同维护股票市场的健康稳定乃至金融稳定。In this paper,an artificial intelligence-based prediction model of Chinese stock prices is constructed,and the reasons for the heterogeneity of the prediction effect are explored.It is theorized that stock prices are mainly affected by interest rates,market behavior,technical indicators and the factors of companies value.The empirical research on the prediction of the representative individual stock prices in Shanghai and Shenzhen Stock Market found that:the predicted value of stock price was consistent with the real value in most of the time,and the average deviation of prediction accuracy of daily data model and monthly data model was within 1.7%and 8%respectively.And there is no obvious change in the robustness test of the interest rate as control variable.Further heterogeneous study demonstrated that stock price,trading volume,valuation,are the theoretical and practical factors influencing the differentiation of stock price prediction effect.Thereby provide reference,suggestion and basis for the policy authorities,the market investment indexes and financial institutions,to maintain the health and stability of stock market and finance.

关 键 词：人工智能 长短期记忆网络LSTM 股票价格预测 异质性 

分 类 号：F830.91[经济管理—金融学]