点击次数：

论文类型：会议论文

发表时间：2016-10-29

收录刊物：EI

摘要：This paper proposes a new grouping mechanism based on Bayesian optimization for particle swarm optimization algorithm (BOPSO). It is aimed to solve the discrete combinatorial optimization problems with high search space. Although the existing evolutionary algorithms can solve this kind of problems effectively through ingenious encoding and decoding mechanism, with the increase of the scale of data, the efficiency of EAs will be limited or even go down. So, we propose a co-evolutionary algorithm with a new grouping mechanism based on Bayesian optimization algorithm (BOA). Due to learning BN (Bayesian network) is also a NP hard problem, we use PSO to generate the BNs and sampling the training data sets to help learning network structures. We apply BOPSO to job shop scheduling problems (JSP) and the flexible JSP (fJSP) which belongs to a typical discrete combinatorial problem to verify the algorithm efficiency. Our experimental results and analysis suggest that BOPSO is a highly competitive optimization algorithm for the discrete combinatorial optimization problems, especially for JSP and fJSP.