Proceedings of the 2nd International Conference on Financial Innovation, FinTech and Information Technology, FFIT 2023, July 7–9, 2023, Chongqing, China

Research Article

Stock Price Prediction Based on Optimized LSTM Model

Download135 downloads
  • @INPROCEEDINGS{10.4108/eai.7-7-2023.2338038,
        author={Wandong  Zhai},
        title={Stock Price Prediction Based on Optimized LSTM Model},
        proceedings={Proceedings of the 2nd International Conference on Financial Innovation, FinTech and Information Technology, FFIT 2023, July 7--9, 2023, Chongqing, China},
        publisher={EAI},
        proceedings_a={FFIT},
        year={2023},
        month={10},
        keywords={lstm; principal component analysis (pca); pearson correlation coefficient; stock price prediction; analysis indicators},
        doi={10.4108/eai.7-7-2023.2338038}
    }
    
  • Wandong Zhai
    Year: 2023
    Stock Price Prediction Based on Optimized LSTM Model
    FFIT
    EAI
    DOI: 10.4108/eai.7-7-2023.2338038
Wandong Zhai1,*
  • 1: Beijing Jiaotong University
*Contact email: 20711032@bjtu.edu.cn

Abstract

The variation trend of stock prices is often disturbed by various factors such as investor sentiment, market sentiment, and corporate performance. The combined effect of these factors leads to stochastic fluctuations in stock price change, which results in its high-noise and unstable characteristics. In responding to the problem that results of stock price prediction of original long short-term memory neural network model (LSTM) are not accurate enough, this study combines the various analytical indicators about stock prices when applying the structure of LSTM neural network. The indicators as KDJ, BOLL, MACD, ARBR, and CR, are used as the training set data together with the basic stock price trading data to expand the training set data volume and change the model parameters for prediction. After initially improving the prediction accuracy, the study uses Pearson correlation coefficient and Principal Component Analysis (PCA) to further improve the model, and an optimized LSTM neural network model (P-I-LSTM) was proposed to improve the accuracy while reducing the amount of data in the training set and improving the training speed.