Diabetes patients must monitor and control their own blood glucose levels with the intention of approximating the blood glucose levels and dynamics of a typical person. This is an onerous task for most diabetics and this paper will survey a handful of ways that these challenges might be practically addressed with Machine Learning. The regressive versions of AdaBoost and SVMs are evaluated for feasability for blood sugar prediction in the context of intermittent data. The SVM and AdaBoost classification methods are also evaluated for their performance in predicting hypoglycemic events, which are harmful if left untreated. Possible applications of these methods are presented along with a discussion of future research topics and the many factors effecting the success of diabetes treatments.