Auction-based deep learning-driven smart agricultural supply chain mechanism

Agricultural product prices are subject to significant fluctuations due to variations in sales cycles, impacting people's quality of life and farmers' income while potentially giving rise to social problems. To address this issue in the smart agriculture supply chain, advancements in technologies like big data and artificial intelligence have made it feasible to predict agricultural product demand and price trends. Among the research methods, deep learning, being highly reliant on data and possessing multi-layered implicit information, has emerged as the most popular research method, exhibiting better data representation and prediction capabilities. In this study, we propose an end-to-end agricultural product price prediction model that combines complementary ensemble empirical mode decomposition (CEEMD) with long short-term memory (LSTM). The prediction generated by this model will provide guidance throughout various supply chain links, including production, harvesting, storage, and logistics. Moreover, the study aims to develop auction mechanisms based on the prediction results to achieve an optimal allocation of agricultural products. To address issues of low computational efficiency and limited benefit distribution in the auction process, this study introduces a deep learning-based iterative bilateral auction algorithm (DL-IDA), which improves upon existing methodologies by leveraging the power of deep learning. Empirical analysis is conducted using the daily price of Fuji apples at the Beijing Xinfadi market. The results demonstrate that the CEEMD-LSTM model proposed in this study exhibits superior performance in predicting agricultural product prices. Furthermore, experimental findings validate the effectiveness and superiority of DL-IDA. Compared to existing iterative bilateral auction algorithms, DL-IDA offers better performance in terms of running time, social welfare, buyer benefits, seller benefits, and broker benefits. Consequently, it can contribute to a transparent pricing mechanism for agricultural products.

Automated summary

This study proposes an agricultural product price prediction model that combines CEEMD and LSTM, which shows superior performance in predicting agricultural product prices. The model can provide guidance for various supply chain links. In addition, the study introduces a DL-IDA algorithm to improve the efficiency and benefit distribution of the auction process. Empirical analysis demonstrates the effectiveness and superiority of the model and algorithm in predicting and auctioning agricultural products.