In order to solve the problems of microarray gene expression data with the characteristic of high dimension and small sample, high redundancy and a lot of noise, this paper proposed a novel model FICS-EKELM, which was built based on the combination FCBF feature selection and ensemble optimized method, for gene expression data classification. In the proposed method, it firstly used fast correlation-based filter method(FCBF) to eliminate the irrelevant features and noise, and chose the discriminate feature subsets. Secondly, bootstrap technology produced many sample training subsets, by means of these subsets, it used the improved crow search algorithm(ICS) to select optimal feature subsets and optimized parameters for kernel extreme learning machine(KELM) synchronously. And then, it constructed ensemble classifiers for target gene data classification, which based on the basic classifiers. Moreover, the model used parallel method on multi-core platform multithreading processor, which used OpenMP, to speed up the search and optimization process. Experiment on six public famous gene datasets shows that the proposed method not only achieves a higher classification performance with less characteristic genes, but also greatly improves the classification accuracy. It proves the validity of the proposed method.

[2019-10-05] Printed Article /doc/2018.04.0248/v2