Enhancing Diagnostic Accuracy of Co-occurring Diabetic and Thyroid Diseases using Machine Learning Techniques

Abstract

Efficient classification methods are crucial for accurately predicting co-occurring diabetic and thyroid diseases, addressing substantial global health challenges. These conditions affect individuals across diverse demographics, including males, females, infants, adolescents, and the elderly. This study employs ML algorithms to forecast co-occurring diabetic and thyroid diseases (DTD). Utilizing a dataset sourced from the UCI Machine Learning Repository, feature selection techniques were applied to identify relevant attributes and optimize predictive accuracy. Seven distinct machine learning algorithms, including ID3, J48, Zero R, Random Forest, Multilayer Perceptron, and Naive Bayes, were employed to classify subjects based on their disease status. Our analysis of various machine learning algorithms for predicting co-occurring diabetic and thyroid diseases (DTD) demonstrates notable differences in their performance. The Random Forest (RF) algorithm outperformed others with a remarkable accuracy rate of 95.123%, showcasing its potential for accurate disease classification. Following closely, the Naïve Bayes algorithm achieved an accuracy of 93.8596%, indicating its effectiveness in this context. Additionally, the ID3 algorithm demonstrated respectable performance with an accuracy of 87.7193%. These findings underscore the significance of employing machine learning methodologies, particularly Random Forest and Naïve Bayes, to enhance diagnostic accuracy and inform treatment strategies for individuals affected by thyroid or diabetic disorders.