The Research And Implementation Of Large Scale Multi-dimensional Network Analysis Model

With the development of information technology and the lower cost of storage device,companies have established a large number of databases for their business and store the vast amounts of data.How to use these data to provide guidance and advice to business decision-making is a difficult problem for decision analysts.Online analytical processing(OLAP)is recognized as an effective solution.OLAP can analyze multi-dimensional data quickly and efficiently,and it can also support decision analysis.After twenty years of research and development,OLAP has been mature and standardized.Many commercial databases and data warehouse systems have implemented OLAP module.In recent years,with the rapid development of emerging field of social network,biological information,multi-information fusion and so on,multi-dimensional heterogeneous networks emerged in practical application and the size of the networks is also increasing.The data analyzed by traditional OLAP is organized by fact tables and dimension tables and there is no correlation between the facts.Using traditional OLAP cannot effectively analyze multi-dimensional networks.Faced with this problem,Graph OLAP technology is gradually developed.This technology improves the information model and use graph cube to substitute traditional data cube.Graph OLAP support analyze graph data in multi-dimensional and multi-granularity ways.However,Graph OLAP research is still in its infancy and the analysis ability of the models is inadequate.Most of the models do not support effective and efficient analyzing multi-dimensional heterogeneous networks and large scale data.Aiming the deficiencies of the existing Graph OLAP model,this paper proposes a new model which can analyze large-scale multi-dimensional heterogeneous network in multi-dimensional and multi-granularity ways.The main research contents are as follows:1.This paper proposes a novel multi-dimensional heterogeneous network information model.Based on this model,this paper defines and studies the 2-meta relation path and the n-meta relation path which is a new way to guide network aggregation.2.This paper design Two-Step Multi-dimensional Heterogeneous Graph Cube(TSMH Graph Cube)and extend the traditional Graph Cube model into two-step cube about Entity Hyper Cube and Dimension Cube.This paper also enriches the operations of Graph OLAP and gives more new semantics for traditional operations,which makes network analysis more diverse.3.For Entity Hyper Cube,this paper design aggregate algorithms and proposed a materialization strategy.For Dimension Cube,the use of hierarchical coding for entity types and dimensions makes that the dimension operations of entities do not need to join dimension tables,which improve the efficiency of dimension operations greatly.4.To support analyzing large scale of data,the algorithms of Graph OLAP operations are implemented by using parallel computing framework.The experiments on large-scale real and synthetic data set verify that the framework can analyze multi-dimensional heterogeneous network effectively and efficiently.