Supporting Customers with Limited Budget in Data Marketplaces

Lima Martins, Denis Mayr; Lechtenbörger, Jens; Vossen, Gottfried

Abstract

As the competitiveness and dynamics of current markets intensify, companies and organizations see opportunities to optimize their strategies and increase their business advantage in data-driven decision-making. This has led to an emergence of data marketplaces, where providers can sell data, while consumers can purchase it. However, the process of acquiring data from a marketplace involves issuing queries with an associated monetary cost, and data consumers often struggle to purchase the targeted data set of appropriate volume and content within their budget. Two issues need to be considered: One is querying itself, which may require API calls, structured queries written in SQL, graph queries written in Neo4J, or any other language framework. Querying is often a stepwise process that starts from generic queries and gets refined as the user learns about the data that results. The other issue is the cost involved, which consists of the price a consumer has to pay for the data and that of processing the various queries. In this paper, the second issue is studied from a computational perspective; in particular, we propose a novel framework for data-purchase support that considers data purchase from a marketplace as a sequence of interactions between the data provider (or the marketplace) and the consumer. This allows us to deal with scenarios in which the consumer has a limited budget, insufficient to embrace the complete data set he or she targets. We formalize the problem setting and the characteristics of available queries offered by the data provider so that efficient (approximation) algorithms can be devised. Our empirical results demonstrate that intelligent algorithms can aid the data consumer with near-optimum solutions that consider her preferences about the queries to be issue to the data provider.