Empirical Evaluation of Different Feature Representations for Social Circles Detection

Abstract

Abstract. Social circles detection is a special case of community detection in social network that is currently attracting a growing interest in the research community. We propose in this paper an empirical evaluation of the multi-assignment clustering method using different feature representation models. We define different vectorial representations from both structural egonet information and user profile features. We study and compare the performance on the available labelled Facebook data from the Kaggle competition on learning social circles in networks. We compare our results with several different baselines.

Feature representation for social circles detection using MAC

“…Finally, MAC is a state-of-the-art technique, having recent and influential publications, such as [17,20], in which it was employed and considered as a baseline method for social circles detection. In this work, we continue the research started in [1] and explore further the possibilities of MAC for this task, investigating novel representations. We defend the fact that this technique still has further potentiality and better results can be obtained.…”

“…In [17,20], some tests are performed with this idea, but using only the node attributes, no information from the graph structure. In [1], MAC was fed with representations of both the graph structure and the node attributes, with successful results. The main aim of this paper is to focus on users' profiles features and to conduct an exhaustive study on how many of them are necessary to obtain the best results.…”

“…Moreover, some techniques specifically designed for social circles detection have been developed recently [17,20]. In this article, we continue the research started in [1], based on multi-assignment clustering (MAC) [29,10]. MAC is a clustering technique not necessarily related to networks or graphs, which permits to assign the same object into several different clusters or, in our case, social circles.…”

“…MAC is a clustering technique not necessarily related to networks or graphs, which permits to assign the same object into several different clusters or, in our case, social circles. In [1], we conducted preliminary experiments using this method, defining several vectorial representations of both structural network information and users' profile information, with the aim to investigate which of these representations provides the best results. In this new study, we expand the experimentation, making variable the number of users' profile features employed.…”

“…In this new study, we expand the experimentation, making variable the number of users' profile features employed. Unlike in [1], we allow for the clustering technique to automatically detect the number of predicted circles, instead of using the number of groundtruth circles. We test our method on two different datasets and evaluate it by means of measures present in the literature.…”