This document reviews machine learning techniques for early prediction and risk minimization of diabetic disease. It discusses how various machine learning algorithms like decision trees, KNN, random forests, and SVM have been applied to diabetes prediction datasets. Accuracy rates of 83.11% to 88.42% were achieved for different algorithms. Feature selection techniques like Pearson correlation were also able to improve some algorithm accuracies further. The document proposes using machine learning systems to better diagnose and care for diabetic patients early on.