Optimizing Fresh Data Sampling and Trading

doi:10.1109/TON.2025.3567065

Optimizing Fresh Data Sampling and Trading

0

PRE

AI

DOI：10.1109/TON.2025.3567065

Original

Original request for help

Save

Abstract

En 中文

Existing works on data trading often overlook the impact of data freshness on its valuation. This paper explores a fresh data market, where a platform offers data with varying freshness levels, such as real-time traffic data, to users who arrive stochastically. We categorize data updates into two types: lightweight (e.g., noise level) and computation-intensive (e.g., traffic images). Initially focusing on lightweight updates, we introduce three pricing policies: uniform, dual, and dynamic. The challenge lies in jointly optimizing the platform’s data sampling and pricing, a complex non-smooth mixed integer programming problem. Nevertheless, we achieve closed-form optimal solutions for all three policies by analyzing a relaxed version of the problem. Our findings reveal the surprising insight that higher data acquisition costs lead the platform to lower uniform data prices due to staler, less valuable data. Our numerical analysis indicates that the optimal dual pricing policy closely matches the dynamic pricing policy in performance and substantially exceeds the uniform pricing, tripling profits in some cases. Extending our work to computation-intensive updates, which require preprocessing, adds extra complexity. We tackle this by applying fractional programming. Numerical results show that profits from optimal uniform and dual pricing closely approach those from dynamic pricing, as the platform can adjust processing time.

Keywords:

Fresh data market

age of information

data sampling and pricing

Optimizing Fresh Data Sampling and Trading

Abstract

Journal

Organization