Implementing Artificial Intelligence and Machine Learning Algorithms for Optimized Crop Management: A Systematic Review on Data-Driven Approach to Enhancing Resource use and Agricultural Sustainability

Implementing Artificial Intelligence and Machine Learning Algorithms for Optimized Crop Management: A Systematic Review on Data-Driven Approach to Enhancing Resource use and Agricultural Sustainability Ugwu Okechukwu Paul-Chima¹*, Ogenyi Fabian C.¹, Alum Esther Ugo¹, Val Hyginus Udoka Eze¹, Mariam Basajja², Ugwu Jovita Nnenna¹, Ugwu Chinyere N.¹, Regina Idu Ejemot-Nwadiaro³˒⁴, Okon Michael Ben¹, Egba Simeon Ikechukwu¹, Uti Daniel Ejim¹ ¹ Department of Publication and Extension, Kampala International University, Uganda² Health Care and Data Management, Leiden University, Netherlands³ Department of Public Health, School of Allied Health Sciences, Kampala International University, Uganda⁴ Directorate of Research, Innovation, Consultancy and Extension (RICE), Kampala International University, Uganda Description Corresponding Author: Ugwu Okechukwu Paul-ChimaEmail: [e-mail]; [e-mail]: 0000-0003-3563-3521 ABSTRACT This systematic review synthesizes evidence from 2013–2023 on the applications of Artificial Intelligence (AI) and Machine Learning (ML) in precision agriculture. Focusing on crop monitoring, yield prediction, and resource optimization, the review highlights that neural networks, decision trees, and deep learning are the most frequently applied algorithms. Deep learning demonstrated the highest predictive accuracy (93%), though interpretability remains a barrier to adoption. Quantitative evidence shows AI adoption can increase crop yields by up to 25%, reduce input costs by 28%, and improve operational efficiency by 40%. Resource-specific benefits included 22% water savings, 28% fertiliser savings, and 35% nitrogen runoff reduction. Smallholder farmers benefited through mobile-based AI systems, with yield and pest management improvements ranging from 15–30%. However, challenges such as poor data quality, high infrastructure costs, lack of digital literacy, and ethical issues (data ownership, algorithmic bias) persist. Integration with blockchain, IoT, and robotics further strengthens AI’s potential in climate-resilient agriculture. The findings confirm AI/ML as transformative tools for sustainable farming, but equitable access requires inclusive policies, explainable AI, and international governance frameworks. Keywords: Artificial Intelligence, Machine Learning, Precision Agriculture, Neural Networks, Crop Yield Prediction, Smart Irrigation, Sustainable Farming. Methodology Databases searched: PubMed, Scopus, Web of Science Search date: December 10, 2023 Period covered: 2013–2023 Inclusion criteria: Peer-reviewed studies applying AI/ML in crop-based agriculture with quantifiable outcomes (yield, efficiency, pest/disease management, sustainability). Exclusion criteria: Non-peer-reviewed sources, algorithm-only development papers without field application, and studies lacking quantitative outcomes. Number of studies included: 95 (out of 1,347 initial records) Data extraction and synthesis followed PRISMA guidelines, with inter-rater agreement (Cohen’s Kappa ≥ 0.75). Risk of bias was assessed using a modified ROBINS-I framework. Both qualitative thematic synthesis and quantitative effect size analyses were performed. Supplementary datasets, coding sheets, and analysis scripts are openly archived at:📂 Zenodo DOI: 10.5281/zenodo.16992983 Major Findings Algorithm performance: Deep learning: highest accuracy (93%) but low interpretability. Decision trees: lower accuracy (82%) but highest adoption due to simplicity and transparency. Neural networks: balanced accuracy (89%) and moderate adoption. Resource efficiency: Smart irrigation: 22% water reduction, 12% yield increase. Precision fertilisation: 28% fertiliser reduction, 35% nitrogen runoff reduction. AI-enabled machinery: 30% labour cost reduction. Sustainability impact: 20–25% reduction in pesticide use. 15–20% increase in soil organic matter. 85% accuracy in biodiversity mapping. Adoption barriers: Data quality and interoperability issues (25% data loss in some systems). Infrastructure gaps (only 35% of rural farms with adequate internet). High costs ($10,000–$50,000 per farm). Ethical concerns around privacy, bias, and data ownership. Conclusions and Recommendations AI and ML technologies provide measurable gains in crop yield, resource efficiency, and environmental sustainability. However, smallholder adoption remains constrained by cost, infrastructure, and equity concerns. Future pathways include: Development of modular, interoperable, and explainable AI tools. Investment in digital infrastructure and rural broadband. Policy frameworks for fair data governance and inclusive adoption. Integration of AI with IoT, robotics, blockchain, and edge computing for scalable solutions. Registration and Availability Protocol registration: Retrospectively registered, openly available with full search strings and coding schemes. Data availability: Supplementary materials, bias assessments, nomenclature, and analysis scripts are openly archived on Zenodo (DOI 10.5281/zenodo.16992983) APPENDIX Appendix A – Search StrategiesOne-line search strings (PubMed, Scopus, Web of Science) with filters and search dates. Appendix B – Data Extraction Sheet (Excel/CSV)Includes bibliographic information, crop/system studied, AI/ML details, dataset size, validation type, evaluation metrics, outcomes, adoption barriers, and funding source for all 95 included studies. Appendix C – Risk of Bias Ratings (Excel/Word)Domain-level risk of bias ratings (adapted ROBINS-I + PROBAST/QUADAS-2/ML reproducibility domains). Appendix D Nomenclature Table Analysis Scripts R scripts (meta-analysis, inter-rater reliability using irr package). Python scripts (visualisations using pandas and matplotlib). FiguresForest plots, summary charts, and risk-of-bias heatmaps. README.md (this file) with instructions for reproducibility. SOFTWARE & VERSIONS Microsoft Excel v.16.80 — data entry, descriptive summaries R v.4.3.2 — meta-analysis (meta v.6.5, metafor v.4.5), inter-rater reliability (irr v.0.84) Python v.3.11 — data processing and figures (pandas v.2.1, matplotlib v.3.8) Data & Code Availability All materials are released under a CC-BY 4.0 license. Users may reuse, adapt, and build upon these materials with attribution. Protocol Registration The review protocol (search strategy, extraction scheme, and analysis plan) has been retrospectively registered at [OSF link — insert once uploaded]. Citation If you use these materials, please cite as: Author(s). Supplementary Materials: Systematic Review of AI/ML Applications in Precision Agriculture (2013–2023). OSF/GitHub/Zenodo.