LLM-Augmented Dynamic Topic Modelling Via Lda for Global Big Data Research Indexed in Scopus (2015-2026)

Big Data scholarship has expanded rapidly over the past decade, yet its longitudinal intellectual architecture remains insufficiently synthesised. This study examines thematic evolution, disciplinary convergence, and citation distribution within global Big Data research published between 2015 and 2026. A corpus of 6,302 peer-reviewed articles was assembled through precision Boolean searches within Scopus. Titles and abstracts were pre-processed using stop-word removal, vocabulary normalisation, and lemmatisation to produce a consistent analytical dataset. Latent Dirichlet Allocation was implemented within a dynamic topic modelling framework using the Gensim Python environment, with AI-supported interpretive validation. Model optimisation incorporated coherence evaluation, perplexity diagnostics, and expert assessment, yielding a five-topic configuration. Publication trajectories, citation intensity, and inter-topic correlations were systematically examined to capture structural and temporal dynamics. Five coherent domains emerged, encompassing Health and Risk Management, Algorithms and Machine Learning, Smart Technology and Services, Knowledge and Library Science, as well as Cloud Computing and Storage. Knowledge and Library Science demonstrated dominance with 1,813 publications, 60,388 citations, and a mean citation rate of 33.31. Findings indicate that Big Data research has achieved substantial methodological coherence, growing trans disciplinary connectivity, and increasing epistemological authority within the global scientific community.