In this paper, we present an algorithm for the localization of nodes of a wireless sensor network using machine learning techniques. We determine the features of the underlying network that heavily influence the localization of nodes in a wireless sensor network. Support vector regression is used on these features to localize the unlocalized nodes of the network. We apply our algorithm to range free localization and range-based localization. The hyperparameters of the support vector machines are tuned for the underlying graph in order to obtain optimal performance. We find that, even with as little as 10% of the nodes designated as anchor nodes, we obtain over 90% localization accuracy for range-based localization, and over 80% accuracy for range-free localization. Results demonstrate that our feature selection for the machine learning technique outperforms various traditional node localization algorithms on random wireless sensor networks for different radii of the network.