Optimized Data Mining Approach For Breast Cancer Prediction Using Fuzzy Theory And Machine Learning

Abstract

Breast Cancer (BC) continues to be a major health challenge for women around the globe due to factors such as delayed diagnosis and difficulties distinguishing between benign and malignant tumors. Mammography lacks accuracy and often requires interpretatively invasive procedures that can lead to complications. The current research aims to develop a hybrid BC prediction model based on fuzzy logic and ensemble Machine Learning (ML) techniques, utilizing the WDBC dataset comprising 569 samples. Treated under the ensemble approach, the proposed methodology consists of parallel processes of fuzzy rule-based risk evaluation and ML classification to compute a fitness score for definitive selection. The model’s interpretability and accuracy were improved through fuzzy rule generation alongside feature selection using PCA. The experimental findings showed that the proposed Fuzzy + Ensemble model surpassed the individual classifiers’ performance, attaining a precision of 93%, recall of 91%, accuracy of 99% and 92% in the F1-score. These values indicate a substantial improvement over conventional methods, such as Decision Tree (86%) and Naïve Bayes (84%). The confusion matrix further confirmed a low false positive and false negative rate. The research signifies a notable advancement in diagnostic decision-making by combining interpretability with predictive power, making it a viable framework for real-time, intelligent BC prognosis systems in clinical environments.