A novel integrated diagnosis method for breast cancer detection

Abstract

The effective detection of breast cancer is particularly essential for recovery and treatment in the initial phases. The existing methods are not successfully diagnosis breast cancer in the initial phases. Thus the initial recognition of breast cancer is expressively a great challenge for health professionals and scientists. To resolve the problem of initial stages recognition of breast cancer, we recommended a machine learning based diagnosis method which will excellently classify the malignant and benign persons. In the designing of our method machine learning model support vector machine has been applied to classify the malignant and benign persons. To increase the classification performances of the method, we used Minimal Redundancy Maximal Relevance and Chi-square algorithms to choose more appropriate features from the breast cancer dataset. The training/testing splitting technique is used for training and testing of the model. Additionally, the performance of the model has been evaluated by performance assessment metrics. The experimental results demonstrated that the classifier support vector machine obtained best classification performance on the selected subset of features as selected by Minimal Redundancy Maximal Relevance feature selection algorithm. The performances of support vector machine on selected features by Chi square feature selection algorithm are low as compared to Minimal Redundancy Maximal Relevance algorithm. From experimental results analysis, we determined that the integrated system based on Minimal Redundancy Maximal Relevance and support vector machine performances are high due to the selection of more suitable features and obtained 99.71% accuracy. According to McNemar’s statistical test the proposed method is more significant then existing methods. Thus, we recommend that the proposed diagnosis method for effective detection of breast cancer.