Research And Implementation Of Top-k Algorithms For Homogeneous Symmetric Publish/Subscribe System

In recent years, with the development of Internet technologies, the application of homogeneous symmetric publish/subscribe system is also becoming more and more popular. Because in the homogeneous symmetric publish/subscribe system, the candidate cycle matching result set is very big, especially in large system, the result set could be in the mass. So it is not convenient for users to use these candidate results. How to recommend k candidate ring matchings effectively is one of the key problems of the homogeneous symmetric publish/subscribe system. In different applications of homogeneous symmetric publish/subscribe system, sometimes cycle matching results can call for scoring function to get evaluation values, then according to the evaluation values,it can get k optimal matching results; sometimes cycle matching results can not. In order to solve the problem, this paper respectively proposes a Top-k query algorithm facing to matching results of not sorted and a Top-k query algorithm facing to matching results of sorted.In order to solve the problem of evaluation values can not be obtained by calling scoring functions. This paper analyses and summaries the basic theory of related work to find related technologies about the solution to the problem and proposes the Top-k algorithm based on k-skyline query facing to matching results of not sorted. At first pub/sub model is extended; and then the paper describes the detail of the algorithm and the basic operation. Finally, the proposal is evaluated in a simulated environment from different sides with the number of subscriptions, the data distribution, the selectivity, and the number of dimensions. As a result, the precision and recall of the algorithm are all above94%.When the evaluation value of matching results can be obtained by calling scoring function, different complexities of scoring function may affect the algorithm. For those high complexity scoring functions, this paper proposes a Top-k algorithm based on high complexity of scoring function facing to matching results of sorted, which is based on that one comparison operation is better than arithmetic operation in time efficiency. The core of this algorithm is to reduce the times of calling the scoring function to get better time efficiency. If the comparison operation on each dimension costs less time than calling a scoring function, then the system may save time. The proposal is evaluated in a simulated environment. The evaluation results show that, they are suitable for homogeneous symmetric publish/subscribe system.