Forecasting the Stock Price of Listed Innovative SMEs Using Machine Learning Methods Based on Bayesian optimization: Evidence from China

Innovative SMEs have had an important impact on the economies of emerging countries in recent years. In particular, the volatility of their share prices is closely related to economic development and investor behaviors. Therefore, this study takes the Chinese market as an example, after constructing 34 determinants that affect the stock price, the RF, DNN, GBDT, and Adaboost models under Bayesian optimization are employed to forecast the next day's closing price of listed innovative SMEs. The number of samples is 78,708 from 337 SMEs listed on the Chinese SSE STAR market, from July 22, 2019, to September 10, 2021 period. The experimental results show the RF and DNN models perform at a better prediction level than the GBDT and Adaboost models, in terms of the evaluation indicators of R-2, RMSE, MAPE, and DA. Then K-fold method and t-tests as robustness checks ensure our experimental results are more reliable and robust.