Implementation of Machine Learning Algorithms in Diabetes Prediction

Nowadays, diabetes is a severe medical condition that affects the human body leading to various complications such as heart disease, kidney problems, vision impairments, nerve issues, difficulties in urination, and numerous other health concerns that can put individuals in critical situations. The development of diabetes can be attributed to factors such as obesity, congenital diabetes, and elevated blood pressure, all of which can contribute to the onset of diabetes mellitus. However, if diabetes is detected in its early stages, it can be effectively managed. To achieve this goal, machine learning algorithms are employed to analyze patient data and predict the likelihood of diabetes. This includes algorithms such as Bernoulli Naive Bayes, decision tree, K-nearest neighbor (with hyperparameter grid search), and support vector machine, which are utilized for this prediction task. Each algorithm produces varying results when compared to the others. However, when comparing these outcomes, it is evident that the K-nearest neighbor with grid search optimization techniques and support vector machine yield the most accurate results. Both algorithms achieve a remarkable accuracy rate of 99.03%, surpassing other classification algorithms and ultimately enhancing the accuracy of diabetes prediction, thereby improving the overall health condition of patients.