Machine Learning Model for Breast Cancer Data Analysis Using Triplet Feature Selection Algorithm

Abstract

The machine learning techniques can be used for clinical investigations in breast cancer diagnosis. The researchers investigated various machine learning algorithms, such as Support Vector Machine, Naïve Bayes, Logistic Regression (LR), Random Forest, Decision Tree and K Nearest Neighbor to diagnose the disease. Early detection of breast cancer cells from the features is essential. Feature selection is the process of reducing the input features to improve the performance of the model. This research aims to increase the accuracy, sensitivity, specificity and to reduce the False Positive Rate (FPR) and False Negative Rate (FNR) by feature selection. The proposed feature selection technique is comprised of two phases: feature grouping and feature selection. In the first phase, feature grouping uses the Pearson correlation techniques to identify the correlation among the features and group the features based on high-, medium- and low- level ranking. In the second phase, Triplet Feature Selection (TFS) method has been proposed to avoid collinearity among the features. In this, the features are selected based on the correlation differences in each subset when satisfying the race condition. Finally, select the features in the triplet group and apply LR classification technique to diagnose the disease. The proposed classifier achieved an accuracy (95.4%), FPR (1%), FNR (4%), sensitivity (97%) and specificity (96%) to detect the benign and malign ones. The effects of TFS feature selection with LR classifier were used and the performance of the proposed framework was compared with the existing feature selection methods and classifiers.