Predicting Stock Price Movement after Disclosure of Corporate Annual Reports: A Case Study of 2021 China CSI 300 Stocks

Fengyu Han,Yue Wang

DOI: https://doi.org/10.48550/arXiv.2206.12528

2022-07-22

Abstract:In the current stock market, computer science and technology are more and more widely used to analyse stocks. Not same as most related machine learning stock price prediction work, this work study the predicting the tendency of the stock price on the second day right after the disclosure of the companies' annual reports. We use a variety of different models, including decision tree, logistic regression, random forest, neural network, prototypical networks. We use two sets of financial indicators (key and expanded) to conduct experiments, these financial indicators are obtained from the EastMoney website disclosed by companies, and finally we find that these models are not well behaved to predict the tendency. In addition, we also filter stocks with ROE greater than 0.15 and net cash ratio greater than 0.9. We conclude that according to the financial indicators based on the just-released annual report of the company, the predictability of the stock price movement on the second day after disclosure is weak, with maximum accuracy about 59.6% and maximum precision about 0.56 on our test set by the random forest classifier, and the stock filtering does not improve the performance. And random forests perform best in general among all these models which conforms to some work's findings.

Statistical Finance,Artificial Intelligence