Stock Index Forecasting Method based on Corporate Financial Statement Data

  

  • Received:2021-06-15 Revised:2021-07-27 Online:2021-07-27

CCML2021+317: 结合公司财务报表数据的股票指数预测方法

王基厚,林培光,周佳倩,李庆涛,张燕,蹇木伟   

  1. 山东财经大学计算机科学与技术学院
  • 通讯作者: 林培光

Abstract: All the market activities of the stock market participants had a comprehensive impact on the changes of the stock market, which made the fluctuation of the stock market full of complexity and became a difficult problem to accurately predict the stock price. Among these activities that affected stock market movements, financial disclosure was an attractive and potentially financially rewarding means of predicting changes in stock indexes. In order to deal with the complex changes of the stock market, proposed a method to forecast the stock index based on the financial statement data disclosed by the company. This method firstly preprocessed the historical data of stock index and the data of corporate financial statements, mainly reduced the dimension of the high-dimensional matrix generated by the data of corporate financial statements, and then used the double-channel Long-Short Term Memory (LSTM) to forecast the normalized data. Experimental results on SSE 50 index and CSI 300 index show that the prediction effect of this method is better than that using only the historical data of stock index.

Key words: stock index prediction, financial statement analysis, data dimension reduction, Long Short-Term Memory, dual channel

摘要: 股票市场参与者的所有市场活动综合影响着股票市场的变化,使股票市场的波动充满复杂性,成为准确预测股票价格的难题。在这些影响股市变化的活动中,财务披露是预测股票指数变化的一种吸引人的且具有潜在财务回报的手段。为了应对股票市场的复杂变化,提出一种结合公司披露的财务报表数据进行股票指数预测的方法。该方法首先对股票指数历史数据和公司财务报表数据进行预处理,主要是对公司财务报表数据生成的高维矩阵进行降维,然后用双通道的长短期记忆网络(LSTM)对归一化后的数据进行预测研究。在上证50指数和沪深300指数上的实验结果表明,该方法的预测效果优于仅使用股票指数历史数据的预测。

关键词: 股票指数预测, 财务报表分析, 数据降维, 长短期记忆网络, 双通道

CLC Number: