Assessing Classical and Evolutionary Preprocessing Approaches for Breast Cancer Diagnosis

The utilization of evolutionary machine learning has demonstrated efficacy in addressing challenges related to medical data mining. Medical data mining is an important branch of data mining that plays a big role in the healthcare industry. The present article introduces novel methodologies for predicting the malignancy of breast masses. The Wisconsin Breast Cancer Dataset (WBCD) and Wisconsin Diagnosis Breast Cancer (WDBC) are utilized as experimental datasets for the proposed approaches. The proposed methodology entails three primary phases, beginning with pre-processing and outliers detection, followed by feature selection, and concluding with modeling through machine learning. Notably, the study employs the interquartile range (IQR), Cook’s distance, Z-Score, and K-means methods to detect and remove outliers. Additionally, the Whale optimization Algorithm (WOA) is utilized as a feature selection technique to identify the most relevant features for modeling. Following feature selection, the Support Vector Machine (SVM) and Gaussian Naive Bayes (GNB) algorithms are employed to develop predictive models based on the selected features. The trained models underwent testing, and the results indicate that the optimal accuracies achieved for each dataset were as follows: an accuracy of 9S.25% for the WDBC dataset and a perfect accuracy of 100% for the WBCD dataset.