Research And Implementation Of Parallel Subgraph Pattern Matching Teichniques Orient To Large Graphs Under Cloud Computing Environments

With the rapid development of information technology,all kinds of unstructured data are also bigger and bigger.And most of the unstructured data can be preserved in graph.So the study of graph is also became an important subject.And in today's world has been into the information age,many things in the real world can be described with data.With the continuous development of computer science and Internet,the amount of data we need to deal with is becoming bigger and bigger.So the research to the problem of data mining on the large image has become a very important subject.At present a lot of graph matching algorithm run on a single matchine,in the process of subgraph matching rely on memory overhead,and need global information as a supportDue to the large amount of data is very large,large or atlas cannot fully loaded into memory,although there has been a figure database mining scheme based on disk is proposed,but the cost of access to graph data and I/O generated by the disk are very expensive,this article explores to cloud computing technology on the big picture and muck graph matching technology.First of all,in this thesis,on the basis of the BSP model,we proposed the vertex-centered label propagation based techniques and the muck map matching techniques.Breakthroughs in the past to data calculation and framework as the center,and based on data according to the message genereated among caculation the news subgraph matching.When computing the subgraph match,based on the system characteristics,as well as implementation of graph data filtering.Secondly,this thesis put forward according to the decomposition technique of pattern graph,through the decomposition of patterning will,at the same time in various parts of the model in parallel matching,and then combine the matching results of each part of the model,can reduce the iterative calculation steps,improve the matching speed.Finally,this thesis presents a model based on weight of patterning of Top-k graph matching technology,based on the characteristics of the topology of the model,the model of each side attached a weight,in the selection of Top-k results,through the two levels of aggregation technology,first selected local Top-k matching results,and then calculate the global Top-k matching results.According to the characteristics of the topology of the model and the model is more close to the matching results can be recommended to the user,also can choose to contain different vertex embedded recommended to the user.All in all,this thesis mainly based on cloud computing technology are studied on the larger view of subgraph matching technology,the characteristics of parallel computing framework for the cloud environment,put forward the corresponding solutions.A large number of experiments verify the accuracy and effectiveness of this scheme.