A QoS-driven Customizable Forecasting Framework for Blockchain Transaction Fee Recommendation

In this paper, we propose the QoS-driven Customizable Forecasting Framework (QCFF) to recommend Decentralized Applications (DApps) a transaction fee (TF) to reduce cost of confirming a transaction in blockchain (BC). QCFF provides a uniform API for DApps to access various TF prediction models according to QoS requirements. In addition, a customized TF prediction model can be easily plugged into QCFF. The design and implementation show that QCFF is feasible and applicable for DApps to acquire a recommended TF from various TF prediction models according to their QoS requirements. According to the preliminary analysis, QCFF can help DApps saving near 3.5 USD per transaction according to the ETH price as 2,800 USD in 5/20/2021.