Reinvigorating sustainability in Internet of Things marketing: Framework for multi-round real-time bidding with game machine learning

Abstract

Auction-based incentive mechanisms can satisfy the heterogeneous demands of both Demand Side Platforms (DSP) and Supply Side Platforms (SSP) in Internet of Things (IoT) marketing. However, DSP platforms often need help with two issues during the auction process: low enthusiasm and unreasonable bidding. To address these problems, we use the second-price sealed auction and propose a framework for multi-round real-time bidding with game machine learning. We introduce a multi-round advertising bidding mechanism incorporating reputation incentive rules to enhance DSP enthusiasm. The aim is to stimulate DSP participation and deter malicious DSP behavior, ensuring fairness and transparency in the bidding process. Subsequently, we design an auction screening model and adopt a multi-round auction format to ensure that only capable and willing advertising demand partners can participate, thus guaranteeing the reasonableness of DSP bids. Furthermore, we design a real-time bidding mechanism to adapt to the dynamic nature of the IoT marketing market. This mechanism transforms the problem of maximizing DSP revenue under budget constraints into a parameter adjustment problem based on a Markov Decision Process. We then utilize the Double Deep Q Network method to obtain the optimal bidding strategy for DSPs. Ultimately, the results demonstrate that our framework improves the final transaction price by 14.71%, increases the expected click-through rate by an average of 19.35%, and reduces the average cost per thousand impressions by 20.34%.