Distinguishing Thorne-Zytkow objects from red supergiants is an extremely difficult task. The first step in this task is identifying large, complete, and homogeneous catalogs of red supergiants that are free from contamination by low mass giants. In the coming year, next-generation space telescopes will observe and resolve young stellar populations in the infrared well beyond the Local Group. The sheer number of stars that will be observed makes the identification of red supergiants using ground-based spectroscopic surveys infeasible. While simple photometric methods can also be used, these too suffer from numerous shortcomings. To fill this gap, machine learning techniques can be applied to classify massive stars. Here we present a Support Vector Machine (SVM) classifier that is capable of broadly classifying evolved massive stars with high precision and accuracy using only broad band infrared photometry. In particular, the classifier can identify red supergiants with >90% completeness and efficiency, while rejecting low mass giants. While these results are promising, the quality of available training data inhibits the performance of our classifier. We conclude with a brief discussion of the future prospects of using machine learning to identify red supergiants.