Multidimensional Heterogeneous Information Network Analysis and Mining

点击翻译按钮获取中文摘要

Today's real systems are mostly made of heterogeneous entities that interact with each other through multiple channels of connectivity. For instance, in social networks, different types of actors interact on different contexts via different online platforms. These interactions continually change and could be unique or multiple, similar or different. To better represent such complex systems, a new class of multidimensional heterogeneous information networks has emerged. These networks contain multiple edges between nodes of different types as well as nodes of the same type. Multidimensional heterogeneous networks provide thus general and flexible modeling of most interconnected systems and many practical scenarios. Multidimensional heterogeneous networks are the sources of big scientific challenges for understanding their structural properties as well as their dynamic processes for predicting their multiscale and multicomponent dynamics. The research program that I propose will deal with the scarcity of algorithmic methods for mining such complex networks and will handle original and challenging problems that give rise to novel information network analysis paradigms. My long-term goal is to devise theories and algorithms for mining massive time-evolving multidimensional heterogeneous information networks. My aim is the creation of efficient algorithms that take into account multidimensional interactions between multi-typed entities, the generated content, the dynamic nature of the network, and the heterogeneity of the data sources. As driven applications, my research team and I will consider the analysis of large-scale dynamic social networks in order to provide better understanding of human-Web interactions and give semantic interpretation and insight into the various roles and groupings in online social networks. Furthermore, we will analyze massive anonymized academic data provided by our university to extract meaningful patterns, discover various groupings, track students' trajectories over time, predict actionable insights, and then utilize this information to elaborate an effective students' retention and success plan. This program deals with the lack of effective approaches for mining complex semantically-rich multidimensional heterogeneous information networks. The outcomes of this research program can thus be important as they will equip data analysts with innovative tools to delve deeply complex systems that include multiple and heterogeneous layers of connectivity. Furthermore, this research program represents a unique opportunity for the training of HQP. The students involved in this research will work on challenging problems and concrete applications which will give them a notable advantage in their publishing efforts and also give them expertise that will be useful to meet the needs of academia and the industry as well.