<![CDATA[Breast cancer continues to be one of the major causes of deaths due to cancer amongst women around the globe, requiring an effective method of diagnosis. In this research, a machine learning model pipeline that uses a hybrid weighted soft-voting based ensemble method on breast tumor classification for binary data is presented using the Wisconsin Diagnostic Breast Cancer Dataset (WDBC). Data preprocessing includes balancing the classes using Synthetic Minority Over-sampling Technique (SMOTE), removing the highly correlated attributes, and reducing dimensionality through Principal Component Analysis (PCA). Using stratified 10-fold cross-validation, an ensemble yielded 99.3% accuracy and 100% recall for the malignant class on PCA-transformed data, which was better than any individual classifier. Logistic Regression had good performance too, with 98.83% accuracy and 99.81% ROC AUC, which shows that our data can be nearly linearly separable. The feature importance analysis showed that “worst concave points” and “mean radius” were the most important features, and this makes sense from a medical perspective. Overall, this work presents an effective methodology for diagnosing breast cancer.]]>
Copyrights © 2026
