Osprey-Optimized Support Vector Machine-Based Colorectal Cancer Diagnosis for Handling Sample Imbalance Problem in Microarray Technique

The predication and classification of colorectal cancer can be achieved using the prevalent method of microarray technique. But this microarray technique suffers from the dimensionality problem when imbalanced samples exists during the process of feature gene selection which in turn results in low prediction accuracy. It is thus essential to construct a vital model that aids in preventing the aforementioned challenges with maximized accuracy during the prediction of colorectal cancer. In this paper, Osprey-Optimized Support Vector Machine-based Colorectal Cancer Diagnosis model is proposed for handling the problem of sample imbalance in microarray technique. It is proposed as an ensemble model that helps in classifying samples into healthy and cancer classes in order to improve the performance of prediction. This implemented model comprises of three functional modules. The first module adopted the benefits of redundancy maximum relevance (mRMR) for eliminating redundant genes that contributes toward reduced feature dimensionality to improve the results of the prediction. The second module the merits of RUSBoost for handling the problem introduced by imbalanced data. The final module utilizes the Osprey Optimization Algorithm (OSPOA) for optimizing the features given as input to the support vector machine (SVM) model for attaining classification of healthy individuals and cancer patients. The experiment conducted for the proposed OSP-SVM confirmed an improved sensitivity of 21.98%, G-Means of 29.56%, and specificity of 24.19%, better than the baseline classifiers used for investigation.