LAUSR

Skyline queries play an important role in multiple criteria decision making problems. A representative skyline set contains $k$ skyline points that can represent its full skyline set, which is more significant. The distance-based representative skyline ( $k$ -DRS) is a kind of representative skyline, which can describe the tradeoffs among different dimensions offered by the full skyline set. The previous works only focus on the $k$ -DRS queries in total order domains. In this paper, we focus on the $k$ -DRS problem in both total and partial order (PO) domains. Since $k$ -DRS is a NP-hard problem in $d$ -dimensional ( $d\ge 3$ ) space, it is impossible to calculate the exact $k$ -DRS in $d$ -dimensional space. By in-depth analyzing the properties of the $k$ -DRS, we propose an algorithm $k$ -DRS query algorithm in total and partial order (DRSTP) to solve the $k$ -DRS problem from a new perspective. In DRSTP, first, a value in PO domain is transformed to numerical values in total order domains. Thus, the distance of any two partial values can be measured using the transformed values. Then, we apply US-ELM to divide the full skyline set into $k$ clusters. Next, in each cluster, we design a method to select a point as the representative point. Experimental results show that our DRSTP significantly outperforms its competitors in terms of both accuracy and efficiency.