This study aims to systematically evaluate machine learning (ML) applications in rural agricultural contexts, a critical yet underrepresented area in the literature. Unlike prior reviews focusing on high-tech farming environments, this research uniquely centers on smallholder and resource-constrained systems to explore the intersection of ML, sustainable agriculture, and rural development. A Systematic Literature Review (SLR) following PRISMA guidelines was performed. After screening (N = 496 articles included), we analyzed publications from 2000 to 2024 (search conducted on 15 June 2025). Quantitative findings show a fourfold increase in publications between 2019 and 2024 and that CNN-based methods were used in approximately 8% of image-based studies (see Results for breakdown). The analytical process encompassed five structured stages: data importation, descriptive analysis, interactive visualization, linkage analysis, and insight extraction, ensuring analytical rigor and replicability. The results reveal that although ML technologies such as CNNs, SVMs, and LSTM networks are increasingly used for crop monitoring, disease detection, and irrigation management, their deployment remains predominantly confined to well-resourced agricultural systems. Rural applications face persistent challenges, including limited digital infrastructure, data scarcity, and low digital literacy. Moreover, digital systems have improved rural education processes, showing potential for broader agricultural applications. This study contributes by identifying methodological trends and context-specific gaps, offering a roadmap for developing adaptable, low-cost ML solutions. This review excluded non-English publications and restricted access content for reproducibility of reported methods, which may bias results towards English-language and open-access outlets. To quantify potential bias, we conducted an exploratory search (not included in the primary dataset) which identified 8% additional non-English records in the same query terms — suggesting a non-trivial contribution of non-English literature. Future reviews should expand to non-English sources and include more extensive searches of regional databases to reduce language bias.
Sukriadi et al. (Mon,) studied this question.