Diabetes mellitus is one of the swiftly increasing metabolic disorder diseases. If this disease is uncontrolled its side effects are very hazardous and can damage various organs of the body. It can dysfunction heart, eyes, kidney, and nerves in the human body. The International Diabetes Federation has estimated that 537 million people worldwide are affected by diabetes mellitus. The traditional technique in diagnosing diabetes mellitus is done through risk factor assessment, clinical judgements, and blood tests. Though this approach operates with its intrinsic strength it deliberates its estimation to foresee this disease. An aforementioned envisaging of this disease facilitates the healthcare professionals to identify the risk factors and develop more effective treatment strategies to reduce the mortality rates. Implementation of machine learning algorithms enables a proactive data driven research in identifying trends to predict the occurrence of this disease. The objective of this study is to constitute a pivotal role in surmising the existence of diabetes mellitus by applying machine learning algorithms on two different diabetes datasets. The first dataset for this examination is done on the PIMA dataset collected from the National Institute of Diabetes and Digestive and Kidney Diseases. The second dataset considered in this research is UCI diabetes dataset. SMOTE (Synthetic Minority Oversampling Technique) is applied on the imbalanced dataset to balance it. SHAP (SHapley Additive exPlanations) is applied to know the importance of the features. Supervised machine learning algorithms such as random forest, support vector machine, decision tree, and XGBoost are used in this study. The outcome of this exploration is to ease the intricacy of healthcare professionals to control the vast spread of this disease and to reduce the mortality rate.