Hybrid Adaptive Threshold Segmentation for predicting Breast Cancer

Abstract

Breast cancer is a leading cause of death for women in many parts of the world. The disease is often misdiagnosed until it has progressed beyond the point of effective treatment. Therefore, early detection of the disease would aid in reducing mortality and other associated risks. Microarray gene expression data is difficult to identify and interpret, making it challenging to evaluate and choose the most relevant set of genes for use as breast cancer markers. Our research utilized Matlab 2018a and the Breast Cancer Wisconsin Diagnostic dataset to develop a mixed machine-learning model for fast breast cancer prediction. Here, we apply various machine learning techniques including random forest classification, logistic regression, support vector machine and Naive Bayes, as well as to a dataset to make predictions about their development and eventual size. The success percentage for the eXtreme Gradient Boosting classifier compared to various machine learning approaches is 99.78%. This new approach to prediction has the potential to revolutionize the detection, analysis and prognosis of breast cancer.