Diabetes is a chronic illness characterized by either insufficient insulin production by the pancreas or ineffective insulin utilization by the body. Blindness, kidney failure, and stroke are among the severe side effects of this disease, which is growing throughout the world. With the potential to save millions of lives globally, diabetes early detection is an important study topic. In order to aid in the prediction of diabetes, we have employed three machine learning techniques in this paper: Random Forest, Support Vector Machine (SVM), and Neural Network (NN). Several pre-processing processes are carried out, such as label encoding, normalization, and null value removal, before the pre-processed data is entered into the ML model for assessment. On the Pima Indian dataset, we evaluated the model’s performance with 98.99% accuracy using Random Forest and 73.59% accuracy using Support Vector Machine. The number of false-negative detections can be greatly decreased by utilizing a balanced dataset, though.