Biological phase separation is a thermodynamically driven behavior that dictates the formation of many critical membraneless bodies in the cell, including the nucleolus and stress granules. Because of this biological significance, the abnormal regulation of this phenomenon has also been implicated in diseases, including neurodegenerative conditions like Huntington's and Parkinson's diseases. Despite an ensuing need to characterize phase separation in systems of interest, current experimental and computational techniques are often costly and time-intensive. To bypass these obstacles, this project applies a computational approach, using support vector machines, to predict phase separation in networks based on their topological features. These models consistently and accurately predict binary phase separation outcomes in both a constructed dataset of networks and real, complex networks. The promise of this approach demonstrates an ability to determine phase separation in a network without the need to perform lengthy experiments or calculations.