Performance Comparison of Selected Swarm Intelligence Algorithms on Breast Cancer Diagnosis

Abstract

Techniques of data mining have a growing reputation in healthcare for the diagnosis of breast cancer due to its robust diagnostic capability and better classification. Analyzing huge features generally require large memory and computation. If features are carefully chosen, it is expected that the features set will remove the relevant information from the input data to perform the desired task using this reduced features representation. This paper conducted a comparative analysis of the performance of selected swarm intelligence algorithms on breast cancer diagnosis. Genetic Algorithm (GA), Particle Swarm Optimization (PSO), HarmonySearch (HS) and Tabu Search were employed to select the most relevant from the Wisconsin breast cancer (original) dataset. The selected features were passed to seven different machine learning algorithms: Support Vector Machine (SVM), Decision Tree (C4.5), Naïve Bayes (NB), K Nearest Neighhood (KNN), Neural Network (NN), Logistic Regression (LR) and Random Forest (RF). The earlier study applied PSO, Imperialist Competitive Algorithm (ICA), Firefly Algorithm (FA) and Invasive Weed Optimization (IWO) on the original Wisconsin breast cancer dataset. The breast cancer diagnostic model was evaluated based on accuracy, precision, recall and F1-measure. The results showed that the best accuracy of 97.1388% was obtained in the PSO and Tabu-search applying RF classifier, the best precision value of 0.9720 was recorded in Tabu-search applying RF classifier, the best kappa statistic value of 0.9372 was obtained in Tabu-search algorithm applying RF classifier, the best recall value of 0.9750 was achieved in PSO using RF classifier, the best F1-measure value of 0.9700 was obtained in HarmonySearch Algorithm applying SVM classifier. Finally, it was revealed that the RF classifier performed efficiently with swarm intelligence algorithms.