Incremental Learning Bayesian Networks for Financial Data Modeling

Discovering underlying relationships among financial variables will strongly support various financial researches. In this paper, A novel incremental learning algorithm for Bayesian networks is proposed to build up the relationships among financial variables automatically. Our algorithm can partially update the learned structure according to the new generated financial data, which provide a realtime guarantee on our algorithm. Experiment results show that our algorithm outperforms all the available incremental learning algorithms, even some widely used batch learning algorithms for Bayesian networks both on classic data sets and real financial data sets.