The advancement in technology and changes in lifestyle is a root cause for the significant decrease in physical activities. A decrease in physical activity produces problems that are linked to an individual’s day-to-day life choices. If an individual’s daily habits are poor, it might lead to chronic noncommunicable diseases. Diabetes is a chronic noncommunicable disease that exposes people at risk. So, early diabetes prediction model can help to reduce the risk level of diabetes. The study proposes an effective Artificial Neural Network (ANN) model with SMOTE and SMOTE-Tomek that can achieve better classification accuracy for diabetes prediction. The study consisted of two fundamental parts. Firstly, the study explored handling the imbalanced data by performing both oversampling and under sampling using SMOTE and SMOTE-Tomek. Secondly, ANN model was employed with and without dropout layer for diabetes prediction. SMOTE and SMOTE-Tomek are two different data augmentation method which performs Oversampling and Under sampling on the dataset to improve the performance of the model by overcoming the imbalance dataset problem. ANN is utilised as a classifier in this study to make accurate predictions on the PIMA Indian Diabetes dataset. Important features are selected by implementing Filter, Wrapper and Embedded methods. The selected features are ranked by their priority. The performance of the ANN model is evaluated with and without dropout layer. Based on the experimental result ANN -SMOTE with dropout layer performed well by scoring accuracy 95.1 and ROC curve 90.4%.