Diabetes is a chronic disease due to high glucose levels in the body. Diabetes is caused due various factors like age, genetic heredity, body fat, bad diet, bad lifestyle, and lack of exercise. Diabetes has become a serious disease because there are reasons for it. First, there is no cure for diabetes and second diabetes in turn can cause other diseases like kidney disease, heart disease, eye complications, nerve damage, and stroke. When patients visit hospitals, their diabetes-related data are collected through various tests by the hospitals. Thus, various hospital contains huge amount of diabetes-related datasets. These data can be used by the machine learning algorithms as raw data to study their pattern and make predictions and analyses. The datasets have different attributes like number of pregnancies, glucose level, blood pressure, body mass index (BMI), age, insulin, Skin thickness, polyuria, polydipsia, polyphagia, and many more. In this paper, we are studying and reviewing various machine learning models developed and proposed by researchers for diabetes prediction. In addition to that, we are also implementing various machine learning algorithms for diabetes prediction and applying root assessment method (RAM) for comparison. It is observed that among all the machine learning algorithms random forest is the most accurate in prediction.