Diabetes mellitus, a prevalent long-term metabolic syndrome, poses a substantial public health challenge worldwide, with escalating prevalence rates projected in the coming years. Detecting and managing diabetes early is crucial for reducing the risk of severe complications linked to the condition. Conventional diagnostic techniques can be cumbersome and may not always pick up on subtle signs of the disease. In recent times, machine learning (ML) methodologies have shown notable promise in enhancing the accuracy of diabetes detection. This research investigates the efficacy of various ML algorithms, in conjunction with an incremental Artificial Neural Network (ANN) model, for the identification of diabetes. The study employs a dataset comprising information from 390 patients and 14 distinct features. Before deploying the models, data preprocessing techniques are employed to address any missing values. Results demonstrate SVM achieving the highest accuracy of 92%, followed by the ANN model with 94% accuracy. The ANN model employs an incremental approach to optimize hidden layers, showcasing superior performance compared to traditional ML algorithms. Future research directions include exploring additional ML models, feature engineering strategies, and integration of diverse data sources to further enhance diabetes detection accuracy. This research adds to the progression of ML-driven diagnostic tools, with the goal of streamlining early intervention and enhancing health results for individuals either at risk of or already diagnosed with diabetes.