SVM-ANN: A Hybrid Algorithm for Breast Cancer Early Detection as Benign and Malignant

Abstract

Breast cancer is a form of malignant tumour that occurs in the mammary glands. Most breast cancers begin in one of two places: the lobules or the ducts. Cancer cases are expected to increase from 14 million to 22 million in the next 2 decades, and this number will continue to grow progressively every year, so early detection of breast cancer is essential as an initial step in treating breast cancer. In the medical procedure, patients undergo a series of tests, namely ultrasound, biopsy, and mammography, based on the variation of breast cancer symptoms experienced. Automatic detection of breast cancer can be achieved through classification using machine learning (ML). This study is divided into two stages for analysis. The first stage will explore 5 ML algorithms, namely KNN, LR, SVM, ANN, and the hybrid ML algorithm, namely SVM-ANN, optimized for classifying breast cancer. The second stage in this study is to apply feature selection using two methods, namely the wrapper and embedded methods, using 4 ML algorithms, and the optimized ML SVM-ANN algorithm, and compare the performance results of each method and determine which method provides the most optimal performance results in classifying breast cancer. In this study, using the SVM-ANN algorithm and the embedded feature selection method with the Ridge Embedded technique, the highest model performance results were achieved, yielding an accuracy of 97.86%, a precision of 96.5%, a recall of 96.5%, and an F1-score of 96%.