Weighted Set Containment in Redundant Data Sets

Abstract

Given a set family F and a query set Q, weighted set containment is the problem of finding the set C 2 F with the largest sum of weights, such that C Q, where every element has a corresponding weight. This problem was investigated at the request of Volvo Group IT, who rely heavily on weighted set containment queries in many of their applications. In this thesis we show that weighted set containment can be solved efficiently using trie based preprocessing when applied to redundant data sets. We show that finding the most efficient trie which represents F is NP-complete and we introduce a number of approximation algorithms. We show through empirical testing that some of our algorithms outperform state-of-the-art methods for similar problems when applied to Volvo’s particular data set.