Incremental learning Bayesian networks for the stock return prediction

A group of hybrid incremental learning algorithms for Bayesian network structures is proposed in this paper. The center idea of these algorithms is combining the polynomial-time constraint-based technique and the search-and-score technique together to reduce the computational complexity. Our algorithms offer considerable computational complexity savings while obtaining better model accuracy compared to the existing incremental algorithms. One of these algorithms is also used to solve the stock return prediction problem which still has no good solutions till now.