概率反Top-k查询上Why-not问题研究

发布时间：2018-07-16 16:24

【摘要】：目前,越来越多的研究关注不确定数据集,譬如市场分析、互联网以及大型传感器网络等。同时,不确定数据集上的排序查询也受到了广泛地研究。其中,概率Top-k查询可以找到用户最感兴趣的k个对象。然而在实际应用中,用户预期的对象可能不出现在查询结果中。因此,他们可以提出一个概率Top-k查询上的Why-not问题。另一方面,概率反Top-k查询一般用于商业决策以帮助商家找到对其产品最感兴趣的消费群体以便更好地进行产品营销。当生产商发现自己的目标人群并没有出现在查询结果中,他们可能想知道:为什么某消费群对自己的产品不感兴趣?尽管Why-not问题已经有很多的相关研究,但由于Why-not问题与查询相关,因而现有的Why-not问题处理方法不能被直接用于解决概率Top-k查询和概率反Top-k查询上Why-not问题。基于上述分析,本文主要研究了两个问题:(i)概率Top-k查询上Why-not问题:本文给出了修改Why-not向量Wm和参数k的方式(MWK)来对原始的概率Top-k查询进行优化,以使得Wm能够包含在优化后的查询结果中;(ii)概率反Top-k查询上Why-not问题:本文提出了一个框架WNPTR。给定一个原始概率反Top-k查询和一个缺失向量集合,WNPTR将采用以下三个不同策略分别为用户找到一个优化后的查询:1)修改查询对象Q,2)修改缺失向量Wm和参数k,3)同时修改查询对象Q、缺失向量Wm以及参数k最后真实与合成数据集上的大量实验验证了本文所提出算法的有效性和高效性。
[Abstract]:At present, more and more researches focus on uncertain data sets, such as market analysis, Internet and large sensor networks. At the same time, sorting queries on uncertain datasets have also been widely studied. Among them, probabilistic Top-k query can find k objects of most interest to the user. However, in practical applications, the objects expected by the user may not appear in the query results. Therefore, they can propose a probabilistic Top-k query on the Why-not problem. On the other hand probabilistic anti-Top-k queries are generally used in business decisions to help businesses find the consumer groups most interested in their products in order to better market their products. When manufacturers find that their target population is not in the search results, they may wonder: why is a consumer not interested in their product? Although there have been many researches on the Why-not problem, the existing Why-not problem can not be directly used to solve the probabilistic Top-k query and probabilistic anti-Top-k query. Based on the above analysis, this paper mainly studies two problems: the Why-not problem on the (i) probabilistic Top-k query. In this paper, the way of modifying Why-not vector Wm and parameter k (MWK) to optimize the original probabilistic Top-k query is given. So that Wm can be included on the; (ii) probabilistic anti-Top-k query in the optimized query results. In this paper, a framework WNPTR is proposed. Given an original probabilistic anti-Top-k query and a set of missing vectors, WNPTR will use the following three different strategies to find an optimized query for the user: 1) modify the query object Qn2) modify the missing vector Wm and the parameter KN3) at the same time A large number of experiments on the query object Q, the missing vector Wm and the parameter k on the real and composite datasets show that the proposed algorithm is effective and efficient.
【学位授予单位】：浙江大学
【学位级别】：硕士
【学位授予年份】：2017
【分类号】：TP311.13

【相似文献】