Analyzing Potential of SVM Based Classifiers for Intelligent and Less Invasive Breast Cancer Prognosis

Accurate and less invasive personalized predictive medicine relieves many breast cancer patients from agonizingly complex surgical treatments, their colossal costs and primarily letting the patient to forgo the morbidity of a treatment that proffers no benefit. Cancer prognosis estimates recurrence of disease and predict survival of patient; hence resulting in improved patient management. Support Vector Machines (SVMs) are shown to be powerful tools for analyzing data sets where there are complicated nonlinear interactions between the input data and the information to be predicted. In this paper, we have targeted this strength of SVMs to analyze the potential of classification through feature vectors for predicting the survival chances of a breast cancer patient. Experiments were performed using different types of SVM algorithms analyzing their classification efficiency using different kernel parameters. SEER breast cancer data set (1973-2003), the most comprehensible source of information on cancer incidence in United States, is considered. Sensitivity, specificity and accuracy parameters along with RoC curves have been used to explain the performance of each SVM algorithm with different kernel types.