Dual Metaheuristic Optimization Algorithms for Breast Cancer Diagnosis Model
Data mining techniques are increasingly applied in the medical field for disease diagnosis and classification, including breast cancer, due to their high diagnostic accuracy. A major challenge in breast cancer diagnosis lies in managing numerous features, which makes efficient feature extraction and selection essential for improving classification performance. This study aims to conduct a comparative analysis of machine learning algorithms using dual meta-heuristic optimization techniques—Firefly and Cuckoo algorithms for effective breast cancer diagnosis. The Wisconsin Breast Cancer Dataset was used as the benchmark dataset. Two feature selection methods, Firefly and Cuckoo optimization algorithms, were employed to identify the most relevant features. These selected features were then classified using six machine learning algorithms: Support Vector Machine (SVM), Decision Tree (C4.5), K-Nearest Neighbors (KNN), Neural Network (NN), Logistic Regression (LR), and Random Forest (RF) with stochastic gradient enhancement. Model performance was evaluated using accuracy, precision, recall, F1-measure, kappa statistics, and model-building time. The experimental results revealed that the highest classification accuracy of 97.74% was achieved with the Cuckoo optimization algorithm combined with the RF classifier. Meanwhile, the best precision, recall, and F1-measure values (all 0.9780) were obtained with the Firefly optimization algorithm coupled with the RF classifier. These findings highlight the effectiveness of meta-heuristic feature selection in improving breast cancer diagnosis.
Download Paper