【Author】
Zong, Xuesen; Hu, Zhiqiang; Xiong, Xiaoyun; Li, Peng; Wang, Jinlong
【Source】INTELLIGENT TECHNOLOGIES FOR INTERACTIVE ENTERTAINMENT, INTETAIN 2021
【Abstract】Aiming at the problem of privacy security of parking data and low generalization performance of parking flow prediction model, a federated parking flow prediction method based on blockchain and IPFS is proposed. In this method, blockchain and IPFS are applied to the federated learning frame-work. Under the condition of ensuring the privacy and security of parking data, blockchain is used to replace the central server of federated learning to aggregate multi-party local models. Through blockchain and IPFS, the model data in the training stage of the parking flow prediction model are stored and synchronized quickly, which improves the generalization performance of the model and further improves the training efficiency of the model. In addition, in order to improve the participation enthusiasm of all participants, an incentive mechanism based on data volume contribution and model performance improvement contribution is designed. The experimental results show that the method can improve the generalization performance of the model and improve the training efficiency of the parking flow prediction model, and provide a reasonable reward allocation.
【Keywords】LSTM; Federated learning; Blockchain; IPFS; Parking flow prediction; Incentive mechanism
【摘要】针对停车数据隐私安全和停车流量预测模型泛化性能低的问题，提出了一种基于区块链和IPFS的联合停车流量预测方法。在这种方法中，区块链和 IPFS 被应用于联邦学习框架。在保证停车数据隐私和安全的情况下，利用区块链代替联邦学习的中央服务器，聚合多方本地模型。通过区块链和IPFS，将停车流预测模型训练阶段的模型数据快速存储和同步，提高了模型的泛化性能，进一步提高了模型的训练效率。此外，为了提高所有参与者的参与积极性，设计了基于数据量贡献和模型性能改进贡献的激励机制。实验结果表明，该方法可以提高模型的泛化性能，提高停车流预测模型的训练效率，并提供合理的奖励分配。
【关键词】长短期记忆法；联邦学习；区块链； IPFS；停车流量预测；激励机制
评论