e cien t progressiv e skyline computation - vldb.pdfVIP

EÆcient Progressive Skyline Computation Kian-Lee Tan Pin-Kwang Eng Beng Chin Ooi Department of Computer Science National University of Singap ore Abstract of the second p oint, users are typically lo oking for sat- iscing answers. Fourth, for the same query, dierent In this pap er, we fo cus on the retrieval of a set users, dictated by their p ersonal preferences, may nd of interesting answers called the skyline from dierent answers app ealing. A p erson may b e willing a database. Given a set of p oints, the skyline to pay a little more to b e closer to the city; another comprises the p oints that are not dominated may b e contented with a cheap er hotel as long as it is by other p oints. A p oint dominates another convenient to go to the city. As such, it is imp ortant p oint if it is as go od or better in al l dimen- for the DBMS to present al l interesting answers that sions and better in at least one dimension. We may fulll a users need. present two novel algorithms, Bitmap and In- Traditionally, the DBMS supp orts these applica- dex, to compute the skyline of a set of p oints. tions by returning all answers that may meet the users Unlike most existing algorithms that require requirement. In our tourist example, if the user sp ec- at least one pass over the dataset to return ies \budget to mean in the range of $120-$200, and the rst interesting p oint