Application of Bayesian Network for Nikkei Stock Return Prediction

This paper describes the stock price return prediction algorithms by using Bayesian network. In the first algorithm, the clustering algorithm transforms the stock price return distribution to the discrete values set. The Bayesian network gives the probabilistic graphical model that represents previous stock price returns and their conditional dependencies via a directed a cyclic graph. The network is applied for the stock price return prediction. The second algorithm uses, in addition to the previous stock price return, the prediction error data of the first algorithm for determining the Bayesian network. Finally, two algorithms are compared with the time-series prediction algorithm in NIKKEI stock return prediction.