Diabetes Detection using Machine Learning

Diabetes is the most common disease worldwide. It typically arises when there is an excess of blood glucose or blood sugar levels. Glucose serves as the primary energy source for your body and is generated within the body from the food you consume, primarily through processes like glycogenolysis and gluconeogenesis. In the United States, around 37.3 million people, constituting approximately 11% of the population, grapple with diabetes, primarily Type 2 diabetes, which accounts for 90% to 95% of all diabetes cases. Globally, about 537 million adults live with diabetes presently, a figure projected to rise to 643 million by 2023 and potentially escalate to 783 million by 2045. Our model utilizes machine learning algorithms to calculate the occurrence of diabetes, based on a dataset encompassing demographic and medical information, alongside the diabetes status of patients categorized as either positive or negative. Leveraging various machine learning techniques proves advantageous in prognosticating outcomes by crafting models from patient-derived datasets. Within the scope of this endeavor, we will employ machine learning classification and ensemble techniques on a dataset to anticipate the onset of diabetes. These methodologies encompass K-Nearest Neighbor, Logistic Regression, Naïve Bayes, and Random Forest. Notably, each model exhibits distinct accuracy rates when juxtaposed with others. The maximum accuracy we attained is 98.6%. The culmination of our project endeavors to present a model that attains superior accuracy, effectively demonstrating its capacity to prognosticate diabetes.