关键词: 众包选择查询, 采样, 选择性

Abstract: Crowdsourced query optimization has attracted significant attention from the database community in recent years. In this paper, it considers the crowdsourced selection query with multiple predicates and leverage human power to find all objects that satisfy every query predicate. A straightforward method enumerates every object and checks whether it satisfies each predicate. The cost of this method is [|R|?n,] where [|R|] is the number of objects and n is the number of predicates. Obviously this method is rather expensive, especially for large datasets or many predicates. It finds that different predicates have different selectivities and if it first verifies a highly selective predicate, it can avoid checking other predicates for objects that do not satisfy the predicate and thus significantly reduce the cost. An important problem is to determine a good predicate order. However it is rather hard to obtain an optimal order. To address this problem, it proposes a sampling-based framework to find a high-quality order. In order to control the cost of order generation, it devises a random-sampling-based selection method by randomly selecting the predicate order. Since the random-based selection randomly selects predicate permutations, which may lead to large cost, it proposes a ?ltering-based algorithm to further reduce the cost. It evaluates the method using real-world datasets on real crowdsourcing platforms. Experimental results indicate that the methods obtain a high-quality predicate order while significantly reducing the monetary cost.