Fundamental Analysis based Neural Network for Stock Movement Prediction

Fundamental Analysis based Neural Network for Stock Movement Prediction Yangjia Zheng, Xia Li*, Junteng Ma, Yuan Chen School of Information Science and Technology, Guangdong University of Foreign Studies, Guangzhou, China {yjzheng,xiali,junteng.ma,yuanchen}@gdufs.edu.cn Abstract Stock movements are influenced not only by historical prices, but also by information outside the market such as social media and news about the stock or related stock. In practice, news or prices of a stock in one day are normally impacted by different days with different weights, and they can influence each other. In terms of this issue, in this paper, we propose a fundamen- tal analysis based neural network for stock movement prediction. First, we propose three new technical indicators based on raw prices according to the finance theory as the basic encode of the prices of each day. Then, we introduce a coattention mechanism to capture the sufficient context information between text and prices across every day within a time window. Based on the mutual promotion and influence of text and price at different times, we obtain more sufficient stock representation. We perform extensive experiments on the real-world StockNet dataset and the experimental results demonstrate the effectiveness of our method. 1 Introduction Stock Movement Prediction aims to predict the future price trend of a stock based on its historical price or related information. Stock price prediction can help investors, ordinary users and companies to predict the stock trend in the future, which has good application value. The high randomness and volatility of the market make the task of Stock Movement Prediction a big challenge (Adam et al., 2016). However, with the development of neural network technology, stock movement prediction has achieved good results in recent years (Nelson et al., 2017; Hu et al., 2018; Xu and Cohen, 2018; Feng et al., 2019a; Sawhney et al., 2020; Tang et al., 2021; Zhao et al., 2022). Based on fundamental and technical analysis, existing methods can be roughly grouped into two categories, namely methods based on price factors only and methods based on price and other factors (e.g., news of the stock.). Nelson et al. (2017) used the LSTM(Hochreiter and Schmidhuber, 1997) network to predict future stock price trends based on historical price and technical analysis indicators. Feng et al. (2019a) used the adversarial training as perturbations to simulate the randomness of price variables, and trained the model to work well with small but intentional perturbations. They also extracted 11 related price features to effectively help the model to predict future changes. According to the Efficient Market Hypothesis (EMH) (Fama, 1970), price signals themselves cannot capture the impact of market accidents and unexpected accidents, while social media texts such as tweets could have a huge impact on the stock market. Based on this idea, different models have been proposed to model relevant news texts to improve the overall performance of stock movement prediction. Hu et al. (2018) proposed to use the hierarchical attention mechanism to predict the trend of stocks based on the sequence of recent related news. …

Content truncated to 500 words