A growing incidence of Type II diabetes worldwide has prompted the medical industry to explore solutions for enhancing their medical technology. The fields of machine learning and deep learning are now being actively researched for the development of intelligent and efficient methods for detecting diabetes. This research thoroughly examines and explores the effects of the most recent machine learning and deep learning methods on the detection and categorization of diabetes. The accessibility of diabetes statistics is noted to be restricted. The databases consist of measures obtained from laboratory-based tests and invasive procedures. To develop an efficient solution that is both cost-effective and high-performing, it is necessary to do research on anthropometric measures and non-invasive examinations. Multiple studies have shown the potential to develop detection models using anthropometric measures and non-invasive medical indications. This research examined the effects of oversampling strategies and data decrease in dimensionality via the selection of features.