Agriculture 4.0 is evolving by combining traditional farming methods with advanced technologies from Industry 4.0, offering farmers new opportunities to improve their agricultural practices. However, implementing Agriculture 4.0 through the incorporation of these innovative technologies and evidence-based strategies is facing challenges. Moreover, data is proven as one of the most significant assets in the era of big data. Analyzing massively distributed agricultural data while ensuring its privacy, security, and scalability concerns is also an important challenge. This chapter presents the state-of-the-art by leveraging the applications of federated learning methods to handle the aforementioned challenges and promote collaborative analysis of distributed agricultural data. In this work, we utilize the publicly accessible Rice Dataset Cammeo and Osmancik, which comprises 3,810 instances, with 2180 instances of Osmancik and 1,630 instances of Cammeo. This study presents a federated learningbased rice variety classification (AgriFedClassifier) framework for analyzing distributed agricultural data while safeguarding the privacy and security of clients’ local data. We simulate the framework with Multilayer Perceptron models at each client, training the models for a fixed number of local epochs using local data and aggregating model updates at the server employing Federated Averaging (FedAvg) and Federated Proximal (FedProx) methods. We evaluate the effectiveness of federated learning techniques on horizontally distributed agricultural data under two scenarios: IID and non-IID datasets. Experimental results demonstrate that in non-IID data distributions with 80% of stragglers (nodes encountering delays), FedProx achieved a classification accuracy of 89.33%, whereas FedAvg achieved only 50% accuracy. The results section presents an analysis of the effectiveness of federated and centralized models. Overall, we observe that FedProx effectively managed data heterogeneity, mitigated delays, and improved efficiency compared to FedAvg.
Keywords: Aggregation methods, Agriculture 4.0, Collaborative machine learning, Distributed machine learning, Federated learning, Federated optimization, IID, Industry 4.0, Non-IID.
