Support Vector Machines classify data by finding an optimal hyperplane that maximizes the separation between data points.