Main Article Content
In recent days, human health is highly affected by cancer disease. For facilitating cancer therapy and diagnosis, it is highly important to identify cancer subtypes. Gene expression data based cancer subtype classification has keys to address fundamental issues related with drug discovery and cancer diagnosis. The research work is designed previously using an Artificial Bee Colony (ABC) with Deep Fuzzy Flexible Neural Forest (DFFNForest) approach for classifying cancer subtype. However, high-dimensionality challenge may rise when using the data directly to the classification of the cancer subtypes between samples. To solve this problem the proposed work is designed using a Binomial probability Distribution based Principal Component Analysis (BDPCA). The proposed cancer diagnosis methodology consists of the various stages like dimension reduction, feature selection and classification. Initially, gene expression datasets are taken as input. In first stage, dimensionality reduction is performed by using the Binomial probability Distribution based Principal Component Analysis (BDPCA). In the second stage, feature selection will perform by using Imperialist Competitive Algorithm (ICA) algorithm for reducing classifier’s miss rate. Then selected feature will be implemented in Deep Fuzzy Flexible Neural Forest (DFFNForest). In DFFNForest, fuzzy is used to update the weight values of the classifier in the cancer subtype prediction. With respect to error, f-measure, recall, precision and accuracy, better performance is achieved by proposed system when compared to available systems as demonstrated in experimental results.