Optimizing Multi-Layer Perceptron using Variable Step Size Firefly Optimization Algorithm for Diabetes Data Classification

Abstract

According to a survey conducted by the International Diabetes Federation, the proportion of people living with diabetes is gradually rising. Diabetes mellitus is a chronic disorder caused by elevated blood sugar levels. For the early diagnosis and treatment of diabetes patients, efficient machine-learning methods are needed. Data Classification is a significant subject in many areas of life, and it is also a very challenging job in data mining. Clinical data mining has recently gained attention in complicated healthcare challenges relying on healthcare datasets.  The principal objective of classification is to classify all data in a given dataset to a certain class label. In the healthcare field, classification is commonly employed in much research articles. A hybrid method for diabetes data classification is suggested by integrating multilayer perceptron with a modified firefly optimization algorithm for diabetes data classification. The performance of the proposed hybrid multilayer perceptron variable step size firefly algorithm is compared with other hybrid models such as the hybrid multilayer perceptron particle swarm optimization algorithm, hybrid multilayer perceptron differential evolution algorithm, and hybrid multilayer perceptron firefly optimization algorithm. The performance of these models is calculated based on accuracy, precision, recall, F1 score, and mean square error. In comparison to other models, the proposed hybrid model produces superior outcomes for diabetes data classification.