Combining Federated Learning and Blockchain for Enhanced Traceability in Agricultural Supply Chains

Abstract

To address these challenges, this research will adopt a federated learning method and implement the use of a blockchain system to implement a traceability system on the agricultural chain. Based on the voluntary collaboration of decentralized datasets and blockchain-based ledgers, the technology is poised to enhance the transparency, effectiveness, and reliability of the agricultural chain. TensorFlow Federated utilized in the development of the federated learning model exhibits cifar-10 acuracy of 0. 92, precision of 0. 88, recall of 0. 94 and F1 score of 0, indicating that the classifier significantly outperforms the baseline random forest classifier. 91. The model is tested for joint training from the distributed data sources and shows relatively reasonable training time (12 hrs) and communication complexity (8. 5 GB). The state of affairs of blockchain transactions show that successful transaction records up to 1800 are registered with various activities such as product registration, quality certification of the product, record of transportation of the products, and storage records of the products etc. The datasets included here include various types of agricultural data that cover activities taking place on the farming fields and in processing plants, distributor’s centers, and retail stores amongst other areas of concern and may cover data such as crop yields, weather updates, and transportation routes among others. Some of the software tools used in the development process include Python that can be used for data preprocessing and analysis; TensorFlow Federated which can be used in the creation of federated learning models; Ethereum can be used in the implementation of blockchain. This approach provides an invaluable roadmap for successfully tackling important demands in agricultural SCM systems and optimizing the success of decision-makers and entities along the food supply chain.