Research On Dynamic Top-k Query Supporting Context-aware

As an effective means of data analysis,Top-k query has a more and more important research status in the database field.The existing Top-k query research focuses on improving query efficiency to respond quickly to user demand.However,there are problems with update efficiency and low quality of query results.In the query process,there is no integrated context-aware and unable to adapt to the dynamic change of moving object attributes due to context changes.This thesis studies the Top-k query technology support context-aware,puts forward the integration of context-aware Top-k query structure and Top-k query algorithm that supports user dynamic scenarios and multi-user preference group Top-k algorithm in the preference similarity context.The main research work is as follows:(1)Integrating context-aware into Top-k queries,analyzed the existing Top-k model,on its basis,proposed CIMT query model.In the CIMT query model,a contextual prefiltering algorithm for data splitting is proposed and two separation criteria are proposed.The algorithm uses the situational factor to replace a primitive target object with the corresponding splitting target object,and the splitting target object is the expression of the original target object in a specific situation.And on the basis of the data division grid,we proposed a TTI Index(Trunk Tree Index)used to support a large number of target locating and insert and delete operations in the process of split.And proposed SRG(Search Record in Grid)algorithm.The appropriate Index structure can meet the high efficiency when query is built.Experimental verify the results of query quality and the efficiency of the context prefiltering algorithm are improved.(2)On the basis of TTI Grid Index,a GID-Top-k algorithm(Grid Index based Dynamic Top-k Computation Algorithm)which supports pruning and dynamic change of target property is proposed.GID-Top-k algorithm based on the grid at the dominant relationship,pruning grid by the positions of the grid in the index and summary information,by judging whether the grid has "k control ability" and the "pruning unit" to determine the index divided free area and the influencing area,and raise the efficiency of pruning.The computational module of the algorithm is responsible for calculating the grid partition and calculating the initial Top-k result.When the data change occurs in the free area,the result set of Top-k query is not affected.When data changes occur in the influencing area,the area and calculation are redrawn.Based on the experiment,the gid-top-k algorithm and the existing Top-k algorithm are compared and analyzed in various conditions,and the correctness and efficiency of the algorithm are verified.(3)For user groups with a preference for repetition,the cost-performance ratio of calculate Top-k results in the affected areas is low for each user.Therefore,a group algorithm based on user preference weight is proposed to combine the users with similar preference and find the same k results for them to reduce the computational cost.Two grouping algorithms are designed,EIG(Equal Interval Group)algorithm And SG(similarity grouping)algorithm,and a comprehensive scoring function is proposed to calculate the target objects in the group.Based on the non-grouping algorithm,the cost of the two grouping algorithms and the improved query efficiency and the user satisfaction of the group are compared.