Research On Preidiction And Evolvution Task Of Edge And Community On Multi-dimensional Heterogeneous Network

Prediction and evolution task is a hot spot in the area of complex networkanalysis. Till now, there are a lot of researches on this topic, most of which arebased on homogeneous network (e.g., networks of friends). However, mostreal-world networks are multi-dimensional heterogeneous network, which meansthat there are different types of link between vertices. For instance, the vertices of acollaboration network are authors, while the edges between these vertices can notonly tell us that two authors co-published an article, but can also tells that whicharea this article is about (e.g., database, data mining, machine learning).Multi-dimensional heterogeneous network can carry more semantic informationthan homogeneous network. Thus the research of prediction and evolution on thiskind of network will give us better results. Unfortunately, only a few peoplerealized this.In this paper, we will study the prediction and evolution task onmulti-dimensional heterogeneous network. There are three parts in this paperaccording to the different selection of research granularity and purpose. These threeparts are link prediction on multi-dimensional heterogeneous network, communityevolution model on multi-dimensional heterogeneous network and communitydetection on multi-dimensional heterogeneous network. In the first part, we proposea supervised learning-based algorithm called MCBLP. Experiments on real dataset(DBLP) validate the effectiveness of this algorithm. In the second part, we proposeda community evolving event-based model called MCE. The experiment resultsproved that MCE model can correctly depict the evolution of community. In thethird part, we propose a supervised learning-based algorithm called MCP. Theexperiment results proved that MCP can well predict the evolution of community.