Diabetes is a chronic serious health condition that occurs when the pancreas is no longer produces insulin, or the human body cannot beneficially use the insulin it produces. Recognizing and predicting it at an early stage is the first step towards preventing its progression. With the advent of information technology and its emergence in the medical and healthcare sector, diabetes cases and symptoms are well documented. Knowledge can be discovered for predictive purposes through machine learning and data mining techniques. This work concentrates on evaluating the dataset through classification analysis by utilizing Decision tree, Adaptive boosting, and K-nearest neighbor's algorithms. Thus, a faster model of predicting diabetes is introduced, where the aim is to develop the best model that derives the conclusion on an early detection of undiagnosed diabetes.