Paper detail

Computing the Relative Value of Spatio-Temporal Data in Wholesale and Retail Data Marketplaces

Spatio-temporal information is used for driving a plethora of intelligent transportation, smart-city, and crowd-sensing applications. Since data is now considered a valuable production factor, data marketplaces have appeared to help individuals and enterprises bring it to market to satisfy the ever-growing demand. In such marketplaces, several sources may need to combine their data in order to meet the requirements of different applications. In this paper we study the problem of estimating the relative value of different spatio-temporal datasets combined in wholesale and retail marketplaces for the purpose of predicting demand in metropolitan areas. Using as case studies large datasets of taxi rides from Chicago and New York, we ask questions such as "When does it make sense for different taxi companies to combine their data?", and "How should different companies be compensated for the data that they share?". We then turn our attention to the even harder problem of establishing the relative value of the data brought to retail marketplaces by individual drivers. Overall, we show that simplistic but popular approaches for estimating the relative value of data, such as using volume, or the ``leave-one-out'' heuristic, are inaccurate. Instead, more complex notions of value from economics and game-theory, such as the Shapley value need to be employed if one wishes to capture the complex effects of mixing different datasets on the accuracy of forecasting algorithms. Applying the Shapley value to large datasets from many sources is, of course, computationally challenging. We resort to structured sampling and manage to compute accurately the importance of thousands of data sources. We show that the relative value of the data held by different taxi companies and drivers may differ substantially, and that its relative ranking may change from district to district within a metropolitan area.