ANFIS-based wrapper model gene selection for cancer classification on microarray gene expression data

This paper proposes a gene selection framework, based on wrapper model with neuro-fuzzy approach for cancer classification. ANFIS as a classifier for selected genes from Particle Swarm Optimization (PSO) or Genetic Algorithm (GA) methods applies on six datasets of microarray gene expression data for different cancers. ANFIS is compared with three other classifiers which are Support Vector Machine (SVM), K-Nearest Neighbour (KNN) and Classification And Regression Trees (CART). ANFIS gives the best results for original data of all the datasets and the predictions for noisy data are adequate in comparison with three others classifiers. ANFIS is best for less number genes, clearly. Besides, good results of ANFIS, it can generate TSK type fuzzy if-then rules which are interpretable.