Breast Cancer Diagnosis Using an Ensemble Transfer Support Vector Machine

Abstract

As the number of breast cancer patients increases and the age of onset is younger, early detection and prevention have become the key to prevention and treatment of breast cancer. At present, many classification or clustering algorithms are used to diagnose breast cancer data. However, these algorithms directly lose the minimum source domain information, resulting in a significant improvement in the recognition rate. Based on this, this paper proposes an ensemble transfer support vector machine (ET-SVM) algorithm based on classic support vector machine (SVM). The algorithm can effectively use the knowledge in the source domain to guide the learning of the target task. The result of a single SVM is usually the local optimal solution. And its performance is unstable and its generalization performance is poor. Therefore, this article introduces an ensemble strategy based on AdaBoost algorithm. Experiments on the Wisconsin breast cancer data set proved that the proposed ET-SVM algorithm can achieve better classification results and good generalization performance.