A dynamic graph query-based data pricing method via multi-armed bandits

Graph query-based data pricing is promising for pricing answer subgraphs in graph databases according to diverse buyer queries. However, existing methods provide limited support for adaptive price updates when the broker must balance transaction revenue with transaction success rate under incomplete buyer-valuation information. To bridge this gap, we propose GQPRM , which includes: (1) a vertex-edge pricing mechanism that prices a first-time queried answer graph through non-negative atomic prices; (2) a multi-armed bandit (MAB)-based strategy that discretizes repeated-query prices into candidate actions and uses historical acceptance/rejection records for exploration and exploitation; (3) an entropy-weighted feedback update that allocates the observed subgraph-level price variation to vertices and edges in a demand-sensitive manner, rather than assuming that binary feedback uniquely identifies a single overpriced element; and (4) a smart contract-based incentive allocation mechanism that records ownership and supports auditable reward allocation between the data broker and data owners. Experiments on three typical datasets demonstrate that GQPRM improves cumulative transaction value and transaction success rate over the static baseline under the evaluated settings.