Research On Multidimensional Analysis Method Of Information Network

With the development of database technology, enterprises establish a lot of databases. How can we translate the data into decision information is an important task. Data warehouse emerges. In data warehouse system, OLAP is a widespread technology that can allow us to analyze data from different angles and granularities. Data warehouse and OLAP are based on multidimensional data model:data cube. Data cube is a powerful model that can support OLAF opelations.With the rapid development of computer technology, amount of graph data become available such as social network、biology network、compound network and so on. This kind of data is called information network. In information network, vertex stands for entity and the edge stands for relationship between entities. Each vertex or edge may have attributes、labels and weights. Information networks are ubiquitous. Examples of information networks are:co-author network, social networks such as Facebook and IMDB actor cooperation network. There are two types of information network according to the number of entities types:homogeneous information network and heterogeneous information network. It is important to analyze this kind of data from different angles and granularities, because information networks contain a lot of information of entities and relationships among different kinds of entities.Traditional data cube model is multidimensional data model which is based on the same kind of entity type. For example, in ROLAP databases, all the tuples stands for the same kind of entities. Each field stands for the property of the entity. These entities are independent. So the data cube cannot solve the problem of multidimensional analysis of information network. The research work about information network has a fast development and achieves many results such as graph cube、graph OLAP and so on. But the analysis ability of current research on homogeneous information network is still inadequate and the research on heterogeneous information network is rare. In this paper, according to the structure characteristics of homogeneous and heterogeneous information network, we proposed corresponding multidimensional analysis model:the simple nested cube and multi-layer nested cube. There is a kind of entity type in homogeneous information network and the existing methods are limited which cannot analyze the relationship deeply. The simple nested cube which is proposed in this paper can analyze the entities and relationship equally. There are more than two kinds of entity types and relationship in heterogeneous information network. Multilayer nested cube can be used to model heterogeneous information network realizing the OLAP operations on heterogeneous information network. The main contributions of this thesis are as follows:1. Propose multidimensional network to describe homogeneous information network vividly. On the basis of multidimensional network, simple nested cube is build. Aim at simple nested cube, corresponding compound OLAP query is proposed. The multidimensional analysis on homogeneous information network is solved.2. Propose multidimensional heterogeneous network to describe heterogeneous information network vividly. On the basis of multidimensional heterogeneous network, multilayer nested cube is build. Multilayer nested cube is extended by two-layer nested cube. Aim at multilayer nested cube, corresponding compound OLAP query is proposed. The multidimensional analysis on heterogeneous information network is solved.3. The data pattern and materialization method of nested cube are proposed. The experiment results on DBLP show that this model is efficient and effective.