The IoMT (Internet of Medical Things) environment is proposed in this study for the implementation of an e-diagnostic system based on machine learning techniques, specifically for the diagnosis of diabetic mellitus (type 2 diabetes). However, the inability of ML apps to reveal internal decision-making processes tends to lead to a lack of confidence, which slows end-user adoption in several medical sectors. The Indian Diabetes Pima dataset will be used in this research on 8 machine learning models, including Salp swarm optimisation with logistic regression, voting classifier, logistic regression, linear discriminant analysis, random forest classifier, ridge classifier, gradient boosting classifier, and Naive Bayes classifier models. The accuracy, precision, F1 score, Kappa, and AUC of each algorithm are evaluated based on its performance. To enhance the model, the decision-making process is evaluated. Results show that the proposed model has the highest accuracy, whereas Nave Bayes has the lowest accuracy.