A hybrid of clustering and quantum genetic algorithm for relevant genes selection for cancer microarray data

In order to efficiently explore and exploit large search space, quantum variant of genetic algorithm has been suggested in literature. It utilizes quantum computing principle and genetic operators. Despite the use of the quantum variant of GA, memory and computation time requirements for high dimensional data like microarrays are huge. In this paper, we propose a hybrid approach, ClusterQGA, that uses clustering to select a small set of non-redundant representative genes and then applies Quantum Genetic Algorithm to determine a minimal set of relevant and non-redundant genes. Also a new fitness function is proposed to reduce number of genes without sacrificing the classification accuracy. The effectiveness of the proposed approach in comparison to existing methods in terms of classification accuracy and number of features has been experimentally established for both binary and multi-class publicly available cancer microarray datasets. The proposed approach reduces the computation time of Quantum Genetic Algorithm for high dimension microarray data.