Classification of people at risk for breast cancer using decision tree algorithm, case study: Suddhavej Hospital, Mahasarakham University, Thailand

Abstract

    This research focuses on evaluating the effectiveness of the Decision Tree Algorithm in classifying breast cancer, as well as investigating the associated risk factors. The study employs medical record data from breast mass patients at Mahasarakham University's Faculty of Medicine, spanning 2010 to 2022. The dataset, post-cleansing, comprises 1,524 records, with 1,343 representing low-risk breast cancer patients and 181 representing high-risk cases. The study indicates that the Decision Tree Algorithms, specifically C4.5, C5.0, and Random Forest, exhibit substantial classification accuracy. However, their area under the ROC curve (AUC) values are relatively low due to insufficient class separation, which stems from class imbalance. The research tackles this issue by employing oversampling to augment the minority class instances and undersampling to reduce the majority class instances. The outcomes reveal that both C4.5 and C5.0 Decision Trees yield comparable results, while Random Forest demonstrates a superior AUC and recall, approximately 15-20% higher than C4.5 and C5.0.