Research On Event Matching Based On Multidumensional Content In Information Distribution System

Content-based information distribution system can efficiently meet people's demand for quick acquisition of effective information in a mass of information.Among them,matching algorithm is the key to complete event matching.There are some problems in the existing matching algorithms,such as insufficient division of subscription space,excessive execution times of invalid matches in event matching,low efficiency of event matching and unbalanced index structure.In view of the above problems,this thesis proposes an event matching algorithm,GEM-Tree,which can greatly reduce the number of invalid matches and improve the efficiency of event matching by fully dividing the subscription space.Firstly,this thesis constructs a mathematical model to express the concepts of subscription,event,constraint matching and event matching.Secondly,select the unique typical attributes for all the subscriptions,preliminarily divide the subscriptions from the two aspects of the typical attributes of subscriptions and the number of constraints of subscriptions,and design the attribute filter table to achieve the purpose of preliminarily filtering a large number of mismatched subscriptions.Thirdly,the attribute range mapping to the plane right triangle,the design of the subscription range matching into the area of the plane right triangle GEM structure,quickly screening the range mismatch subscription;Finally,based on the property filter table and GEM structure,GEM-Tree index structure is designed to manage and maintain the subscription,and event matching is completed by GEM-Tree to improve the efficiency of event matching.Based on GEM-Tree standard serial algorithm,the parallel algorithm is designed in this thesis.Using the independence of subscription in the information distribution system,a parallel processing mode of multi-attribute simultaneous matching and multi-range subspace simultaneous matching is designed.Through multi-thread parallel execution flow,parallel event matching is performed on multiple grids in attribute filtering table and multiple cells in GEM structure,thus further improving the efficiency of event matching.In order to fully evaluate the performance of GEM-Tree,this thesis builds an experimental platform with a subscription volume of millions,and compares the performance of GEM-Tree algorithm with the latest BE-Tree,REIN,AP-Index and OP-Index algorithms in various scenarios.In a variety of scenarios where the number of subscriptions,the number of subscription predicates,the size of attribute space and the subscription matching degree were taken as evaluation parameters,the average matching time of GEM-Tree was 2.74ms,2.06ms,3.67ms and 2.02ms,which was 94%higher than that of BE-Tree.GEM-Tree is 4.03 times faster than AP-Index,8.78 times faster than OP-Index and 11.67 times faster than REIN.Experimental results show that GEM-Tree performance is significantly better than the other four algorithms.