Trading Fresh Data with Correlation

The increasing reliance on fresh data in real-time applications underscores the significance of commoditized fresh data. However, current research often neglects the crucial data correlation, essential in applications like intelligent transportation. This paper examines the trading of correlated fresh data, where a platform monitors the time-varying numerical status of multiple correlated data sources. Data users arrive stochastically, each seeking to obtain data from the platform to estimate the real-time status of a specific source of interest. To facilitate data trading, we propose a dynamic pricing policy that allows the platform to adjust prices in real time. We demonstrate that dynamic pricing is an NP-hard mixed integer programming problem and propose an approximate algorithm. Our approach begins with threshold-based data allocation and uses linear programming to optimize pricing, achieving a logarithmic approximation ratio. For binary data sources, we derive an optimal closed-form solution, revealing that data correlation can benefit both the platform and users by offsetting data aging with spatially correlated fresher data. Interestingly, despite users placing a higher valuation on fresher data, the presence of correlation results in fresher data being priced lower. This counterintuitive pricing strategy is designed to encourage users to engage in crosssource data purchasing. Numerical results show that the proposed approximate dynamic pricing policy can achieve at least 90 % of the maximum dynamic pricing revenue. Additionally, data correlation can amplify the platform's revenue by up to 100 % compared to scenarios without data correlation.