Predicting Chinese stock prices using convertible bond: an evidence based on neural network approach

Abstract

Finding the most accurate model for predicting the stock price is a challenging task that academics and researchers must undertake. This study aims to predict the Chinese stock prices, which separate into two groups: high-market value stocks and low-market value stocks. This study uses and compares two models: the Back-propagation Neural Network (BPNN) model and the Extreme Learning Machine Neural Networks (ELMNN) model. In addition, the price of convertible bonds is also considered in this study as a new input variable in the model. This is to determine the optimal model for the Chinese stock data. The prediction results demonstrate that the ELMNN model outperforms the BPNN model regardless of whether the convertible bond price is included as an input variable. Moreover, we discover that the ANNs model performs significantly better for stocks with low market values. In conclusion, the price of convertible bonds has the power to predict stock prices in the future using ANNs models.