Παρουσίαση Πτυχιακής εργασίας Μαρίας Παυλοπούλου, φοιτήτρια Τμήματος Πληροφορικής και Τηλεπικοινωνιών του ΕΚΠΑ, με θέμα: Predicting the Evolution of Communities in Dynamic Social Networks
In recent years, online social networks have exploded in popularity, and their dynamic behaviour is of special interest. Finding patterns of interaction and predicting the evolution of social network communities is beneficial from many aspects. It can contribute to predicting the spread of diseases, information, new trends, ideas and views, as well as to making recommendations to the community members.
In this thesis, we study the problem of community evolution prediction over time in dynamic social networks and formulate it as a supervised learning task using classifiers, where communities are represented as graphs. Structural features derived from the community graph, as well as temporal features describing the past form of a community, are considered as the features that are involved in the task of community evolution prediction. The evolutionary events we try to predict are the continuation, shrinkage, growth and dissolution of communities.
We have experimented with real-life social network data from Twitter and Mathematics Stack Exchange, where we compare the predicted evolutionary events to those of the ground truth.
Twitter post: https://twitter.com/iit_demokritos/status/846628228517302272