Predicting breast cancer risk: A data mining approach

Abstract

Early prediction of breast cancer plays a critical role in successful treatment and saving lives of thousands of patients every year. Although massive clinical data related to the patients is being collected and stored by healthcare organizations, only a small subset of the predictive factors has been used in predicting outcomes. Most of the existing approaches focus on applying statistical techniques on small set of attributes recommended by the domain-experts disease diagnostics. These conventional approaches usually make unrealistic assumptions, e.g., normality, independence or linearity relationships, which may not be always true in practical data. On the other hand, advanced statistical approaches may address some of the above shortcomings; however, they are computationally expensive and may not applicable to massive datasets. In this study, we use a data mining approach which offers significant advantages over conventional techniques to address the existing limitations. Our data-driven approach can efficiently process clinical dataset to discover patterns and reveal hidden information for early detection and successfully treatment of breast cancer patients.