Three-Fast-Inter Incremental Association Markov Blanket learning algorithm

doi:10.1016/j.patrec.2019.02.002

2019-05-01

PRE

DOI：10.1016/j.patrec.2019.02.002

原文链接

原文求助

摘要

En 中文

The Markov blanket is a crucial concept in Bayesian network, and a useful tool for Bayesian network structure learning and causal feature selection. Finding an efficient Markov blanket discovery method is one of the core issues in machining learning, and has been widely studied. The Markov blanket discovery methods are mainly divided into two categories: Non-topology based and topology based. Topology based method is data-efficient, but not time-efficient, while non-topology based method is time-efficient, but not data-efficient. Due to the wide existence of the high-dimensional data with limited samples in practical applications, especially in the medical and biological study with expensive sample collection, we combine the advantages of Inter Incremental Association Markov blanket (Inter-IAMB) and Fast Incremental Association Markov blanket (Fast-IAMB), and propose the Three-Fast-Inter Incremental Association Markov blanket learning (Fit-IAMB) algorithm. Experiments showed that Fit-IAMB had comparable accuracy and much better efficiency than state-of-the-art algorithms, especially for the high-dimensional data with limited samples. (C) 2019 Elsevier B.V. All rights reserved.

Keyword:

Markov blanket

IAMB

Bayesian network