中文摘要：

考虑偏好top-k计算问题,提出一种整合网格索引和DG索引的GriddedDominantGraph（GDG）混合索引结构.首先,提出基于数据点逆支配点集性质的剪枝自由点方法,该方法大大减少了构建索引中的数据点及查询时可能访问的数据点.通过网格索引高效地计算逆支配点集,并得出网格中＂k-最大运算区域＂和＂k-最大查找区域＂,分别在建立索引和top-k查询阶段近似地剪枝自由点.然后,分析了查询索引阶段层次式索引（如dominantgraph（DG））在同一层次中无序访问数据点的不足,通过增加网格索引而使访问有序.计算网格概要信息并将网格单元按函数分值排序,使层次内数据点依据网格单元顺序而访问有序.由于附加的网格索引增加计算和存储开销较少,同时性能有较大提升,所以GDG适用性强.理论分析和实验结果均验证了上述方法的有效性.

英文摘要：

To address the problem of top-k queries with preferable function f（）,this paper proposes a hybrid index structure gridded dominant graph（GDG）which integrates grid index into dominant graph（DG）index.Firstly,it proposes an approach to prune a large amount of free points of top-k queries by reverse dominant point set（RDPS）in constructing and traveling the index of data set.GDG uses grid index to calculate all RDPSes approximately and efficiently.When constructing GDG,grid index figures out k-max calculating region to prune free points that all top-k fucntion f（）would not visit,which decreases the amount of index dramatically.When traveling GDG,grid index figures out k-max search region by f（）which avoids traveling those free points of ad-hoc top-k function f（）.Secondly,it analyses the drawback of layer-based indice,such as DG,which can not rank those data points in the same layer.So,GDG uses grid index to rank those data points in the same layer approximately by k-max search region by f（）,and makes less visited points than DG.Howerver,grid index needs less additional computation and storage that makes the GDG index more adaptive for top-k queries.Analytical and experimental evidences show the efficiency of the proposed approaches.