Diagnosis of Diabetes using Clinical Features: An Analysis based on Machine Learning Techniques

Diabetes, the most frequent and ever-evolving disease, shortens the lives of many people of all ages each year. The high incidence of this illness underscores the need of an early diagnosis. Diabetes is manageable. However, delaying treatment may be hazardous and may end in the patient’s premature death. In addition, diabetes is a costly condition to cure, therefore detecting it early might aid patients by informing them when to seek treatment and how to prepare emotionally and financially. This study’s major objective is to enable physicians in diagnosing diabetic complications in a timely way and combating their increasing prevalence. The dataset including information on diabetes complications has been analyzed using the classification techniques Random Forest (RF), Multi-Layer Perceptron (MLP), and Naive Bayes (NB) to estimate the accuracy of a prediction. To evaluate the efficacy of the different classifiers, we have computed the metrics Accuracy, Precision, Recall, and F1-Score. This study revealed that among all the techniques, Random Forest performed the best with 97.5 % accuracy. It received a high level of accuracy and performed very well on all metric tests, outperforming other leading-edge work.