Numerical experimentation on structure simplification in Bayesian network
Masakazu Shiba, Atsushi Takahashi, Shingo Aoki, Hiroshi Tsuji, Shuki Inoue · 2009
In order to use Bayesian network for a complex world, this paper proposes a framework to simplify the network structure. The framework consists of ¿elimination¿, ¿segmentation¿ and ¿aggregation¿. To verify the framework, this paper discusses the prediction problem on consumer's preferences for the future products. The numerical experimentation shows how the framework could be applied and how it resolved the limitation of Bayesian network for the complex world.