Effect of Diversification and Intensification Trade-Off in Anemia Medical Data Classification

Abstract

Diversification intensification balancing is the major role of metaheuristic algorithms. Diversification (exploration) performs well with evolutionary metaheuristic algorithms, whereas trajectory algorithms are considered the best in local search (explanation). In this work, hybridization between differential evolution algorithm (DEA) and simulated annealing (SA) was conducted to enhance the artificial neural network (ANN) for anemia medical dataset classification. The proposed methodology begins with choosing the best ANN structure (best number of hidden layers and best number of neurons in each layer) after tackling with the hybridization of DEA and SA to obtain an improved classification accuracy. The proposed methodology registered a significant enhancement in anemia classification comparing with four other metaheuristic algorithms; the anemia data was composed of a real dataset gathered from Iraqi blood laboratories to detect anemia diseases. Meanwhile, the proposal applied two benchmarks from the University of California (UCI) repository, namely, Pima Indian diabetes data and liver disorder diseases. To verify the proposed method, we used four metaheuristic algorithms to test the selected medical benchmarks. The metaheuristic algorithms included two trajectory algorithms, namely, simulated annealing SA and Tabu search TS, and two evolutionary algorithms, namely, genetic algorithms GA and differential evolution DE. The proposed method attained remarkable results