Abstract: Complex networks representing social interactions, brain activities, molecular structures have been studied widely to be able to understand and predict their characteristics as graphs. In this study, various real-world networks have been classified according to random graph models that represent them the best. Synthetic graphs generated by the random graph models are used in order to increase the success rate of the classification. It is observed that using higher order graph features, such as 4-motifs and 5-motifs, yields more accurate results. The most distinctive graph features in the classification process are determined by making use of various machine learning algorithms and statistical tools. With the use of different classifier algorithms, our framework is shown to provide higher accuracy and more robust performance.
Authors: Ali Baran Tasdemir, Barkin Atasay and Lale Ozkahya (Hacettepe University)