Bibliometric and Machine Learning-Based Forecasting of Research Trends in Agricultural Hydrogen Energy and Efficiency

This study examines energy efficiency in hydrogen production from agricultural resources through an integrated bibliometric and machine learning framework. A corpus of 958 Web of Science–indexed publications published between 2000 and 2024 was analyzed using Bibliometrix to characterize publication dynamics, citation patterns, journal concentrations, international collaboration, and thematic evolution. In parallel, two long short-term memory (LSTM) models were developed in MATLAB: one to forecast annual publication counts for 2021–2034 and another to generate prospective research keywords from bibliometric text data. The results indicate sustained growth in scientific output, with a marked acceleration after 2015. China and the United States account for the largest citation volumes, whereas International Journal of Hydrogen Energy and Bioresource Technology emerge as the most influential outlets. Recurrent themes include biomass conversion, hydrogen production efficiency, photocatalysis, and performance optimization. The publication-forecast model projects a steep increase in papers through 2028, followed by stabilization above 500 publications per year, while the keyword model consistently highlights evaluation, gasification, and catalytic processing as emerging directions. Overall, the study provides a data-driven map of the field and a replicable approach for anticipating future research priorities at the energy–agriculture interface.