Comparison of Machine Learning Algorithms for Predicting Thyroid Disorders in Diabetic Patients

Abstract

Machine Learning (ML), a subfield of Artificial Intelligence (AI), has been used successfully in the healthcare domain for disease diagnosis. Thyroid disorders and diabetes are two of the most prevalent and interconnected chronic diseases, as both play critical roles in regulating various physiological processes in the body. This study aims to predict thyroid disorders in diabetes patients using six machine learning algorithms: Random Forest (RF), Decision Tree (DT), K-Nearest Neighbors (KNN), Logistic Regression (LR), Naïve Bayes (NB), and Support Vector Machine (SVM). A locally sourced dataset comprising 44,534 instances of diabetic patients was utilized, undergoing preprocessing steps including data cleaning, encoding, and balancing. Two balancing techniques were employed: manual balancing and RandomUnderSampler. The dataset was partitioned into training and testing sets using a Stratified K-Fold cross-validation approach with 10 folds to ensure robust evaluation. Each algorithm’s performance was assessed using metrics such as accuracy and F1-score. Among the models, the RF algorithm outperformed the others, achieving the highest accuracy of 95% on the manually balanced dataset and 84% when the RandomUnderSampler technique was employed. Additionally, the F1-scores for RF were 95% and 82%, respectively, indicating its robustness in handling imbalanced datasets. This study highlights the importance of selecting appropriate preprocessing techniques and machine learning methods for healthcare datasets. The findings can assist healthcare providers in making early diagnoses and interventions for thyroid disorders in diabetic patients, potentially improving their quality of life and overall healthcare outcomes.