EFFICIENT SKYLINE SYSTEM DEVELOPMENT FOR NORMAL AND HIDDEN DATABASES: APPLICATION FOR GOOGLE FLIGHTS

Abstract

Deep web databases provide strict search interface and limited web access with top-k results based on a pre-defined ranking function. However, top-k results may not be suitable for multi-criteria decision making because of the variety in preferences. To make the results more relevant to such a decision maker, skyline records were introduced, and as per definition these records are not dominated by any other record such that a record dominates another if it is better or as good as other for all attributes and better in at least one attribute. In this report, we introduce an algorithm for discovering skyline records from hidden databases using different multi-objective attributes on a real-world database. We predicted a new lower bound for the minimum issued number of queries to extract the skyline. This was supported by our algorithm which accomplished the above task in an efficient manner including the worst-case scenario hence proving our theory via running rigorous experiments on a hidden database given the limitations on hand.