|
چکیده
|
Breast cancer and its recurrence are a major global health issue, impacting a considerable percentage of women over their lifetimes. Accurate prediction of breast cancer recurrence is crucial for enhancing patient outcomes, facilitating prompt interventions, and customizing treatment options. Although machine learning algorithms hold significant promise for breast cancer prediction, there is a shortage of studies dedicated to predicting breast cancer recurrence through these methods; yet, the accuracy of current methodologies remains problematic. In contrast, contemporary research primarily focuses on enhancing prediction algorithms and machine learning models, with inadequate exploration of the importance of intricate feature relationships. This study utilized correlation approaches to generate a graph from the existing breast cancer recurrence dataset, facilitating the extraction of novel features. This led to an expansion of the feature collection based on their correlations, thus enhancing prediction accuracy. This study utilized the Wisconsin Diagnostic Breast Cancer (WDBC) and Wisconsin Prognostic Breast Cancer (WPBC) datasets to examine feature correlations. Four correlation methodologies were evaluated: Pearson, Spearman rank, Kendall Tau, and Point-Biserial. Machine learning methods, such as Support Vector Machines (SVM), K-Nearest Neighbors (KNN), Decision Tree (DT), and Random Forest (RF), are utilized to predict breast cancer recurrence. The findings indicated that the integration of graph-based feature associations significantly enhanced the prediction of breast cancer recurrence, with the Spearman rank correlation and SVM model achieving the highest level of precision.
|