A Framework for Optimized Diabetes Detection Model Based on Binary Butterfly and Machine Learning Algorithms

Abstract

Diabetes has become a major cause of death in both developed and developing countries, affecting a large number of people globally. prompting significant investments in research to find a cure for this critical disease. Traditional approaches reliant on diabetes detection are time-consuming, this necessitates a paradigm shift towards more efficient methodologies. In response, this study introduces a conceptual framework for diabetes detection by leveraging the power of optimized machine learning algorithms. Addressing data preprocessing techniques and optimized feature selection algorithms, and machine learning algorithms, specifically Random forest, multilayer perceptron, and Gradient boosting model, the result shows that Random forest emerges as the potent model showcasing a remarkable performance metrics: accuracy score of 97.66%, F1-score of 97.56%, AUC-ROC of 98.54%, Multilayer perceptron achieved an accuracy of 96.10%, F1-score of 95.96%, AUC-ROC of 98.65% Gradient boost achieved and accuracy of 91.82%, F1-score of 91.49% and AUC-ROC of 98.01% respectively. These findings underscore the significant role of feature selection and machine learning in detecting diabetes offering transformative possibilities for global healthcare enhancement.