Learning ensembles of Continuous Bayesian Networks: An application to rainfall prediction
Scott Hellman, Amy McGovern, Ming Xue · 2012
We introduce Ensembled Continuous Bayesian Networks (ECBN), an ensemble approach to learning salient dependence relationships and to predicting values for continuous data. By training individual Bayesian networks on both a subset of the data (bagging) and a subset of the attributes in the data (randomization), ECBN produces models for continuous domains that can be used to identify important variables in a dataset and to identify relationships between those variables. We use linear Gaussian distributions within our ensembles, providing efficient network-level inference. By ensembling these networks, we are able to represent nonlinear relationships. We empirically demonstrate that ECBN outperforms the meteorological forecast on a rainfall prediction task across the United States, and performs comparably to results reported for Random Forests.