Enhancing Association Rules using Generative Adversarial Networks for Breast Cancer Classification

Abstract

The core algorithms of data mining (DM) enable the discovery of new information and insights by analyzing large amounts of data. Association rules mining (ARM), one of the several DM approaches, is extremely important in DM research. By utilizing ARM in medical diagnosis, early disease detection can be enhanced, and treatment recommendations can be improved based on data-driven insights. Breast cancer remains the leading cause of cancer-related deaths among women on a global scale. It is a huge challenge to researchers in the medical field concerning its diagnosis and prognosis. This paper aims to leverage ARM for the generation of associations that contribute to either recurrence or no-recurrence events in breast cancer. The study utilizes the Breast Cancer dataset from the UCI repository. To ensure comprehensive coverage of associations in both classes, the dataset is balanced using Synthetic Minority Over-sampling Technique (SMOTE) and Generative Adversarial Networks (GAN). Utilizing GAN to balance the dataset enhanced the performance of the association classification.