| With the rapid development of internet and the increasing of computer applications, a large number of graph data have been generated.The fast development of information technology has greatly reduced the cost of storage devices. Based on the need of business requirement enterprises have built a large number of database and stored massive data.The research of using massive data to guide business decision making has become the focus of decision maker. And Online Analytical Processing(OLAP) has evolved into an effective solution to this problem.OLAP is the core technology of achieving business intelligence based on relational data. It supports fast and efficient analysis of massive data with multi dimension and multi granularity and provides decision support.After more than 20 years’ evolution from theoretical research to practical tools, products and business applications, OLAP technology has grown to be relatively standardized and mature and it can be applied to many commercial databases and data warehouses.Recently with the continuous development of social networks,biological networks and compound networks, there are a large number of multi-attribute graph data. Traditional OLAP analysis techniques need to fact tables and dimension tables. The fact tables contain all entities and the dimension tables contain various dimensional attributes of entities.However, the connections between entities have been lost. On this basis,traditional OLAP can not effectively analyze multi-dimensional network.To solve this problem, Graph OLAP technology came into being. In graph OLAP we use graph cube instead of data cube to improve the informational model so that the model can support multi-dimensional and multi-granularity analysis of network data. However, the research of Graph OLAP is still in the initial stage and many models are still in the exploratory stage. And they are inadequate in supporting the effective and efficient analysis of large scale graph data and multi-dimensional heterogeneous network. Aiming at the deficiency of the existing Graph OLAP models, this paper proposes a new analytical model which can support effective and efficient analysis of large scale heterogeneous network. The main contents of this paper are as follows:1. This paper introduces the concept of relation meta path in heterogeneous networks and studies the relationships among meta paths to guide the aggregation of multidimensional networks.2. This paper defines a new graph cube model Star Graph in which introduces the concept of main node to guide the generation of meta path and guides the OLAP operation by the meta path of network. According to the characteristics of the cube model, the operation of Graph OLAP is enriched which makes the network analysis more diverse.3. This paper proposes some parallel aggregation algorithm and the corresponding materialized strategy aiming at the new graph model. Node encoding algorithm is proposed to encode node and node attributes which eliminates the connection operation between entity tables and dimension tables and it greatly improves the efficiency of OLAP operations.4. This paper uses parallel computing framework to support operation of massive data. Experiments on large scale real datasets show that the model is effective and efficient for large scale data analysis. |
