Exploring current trends in agricultural commodities forecasting methods through text mining: Developments in...

Agriculture stands as one of the major economic pillars worldwide, with food production contributing significantly to income growth. However, agricultural activities also entail risks associated with uncontrollable factors along the supply chain. To address these challenges, mathematical models have been developed for forecasting crucial variables in managing agribusiness activities. In this context, this article employs a combination of systematic bibliometric analysis and the Latent Dirichlet Allocation (LDA) method, a semi-automated approach. The main objective of this study was to automate the identification of relevant topics and construct a bibliographic portfolio (BP) covering the period 2015-2022, focusing on methodologies used in articles and other bibliometric analyses. The 30 articles included in the BP address issues related to methodologies applied in the temporal analysis of agricultural commodities. These articles were categorized based on the nature of the prediction models used, classified as (i) machine learning (ML), (ii) machine learning and artificial neural networks (ML-NN), (iii) machine learning and ensemble (ML-Ensemble), (iv) machine learning and hybrid (ML-hybrid), and (v) statistical. Regarding the results, the topic that stood out the most was termed "Forecasting Methods Applied to Agribusiness Time Series." The most utilized classes were ML-hybrid (41.95 %) and statistical (29.31 %), followed by ML-NN (14.94 %), ML (9.20 %), and ML-Ensemble (4.60 %) types. The theoretical contribution of this study lies in identifying literary gaps concerning forecasting methods applied to agribusiness, while its practical implication is to identify forecasting methodologies to support decision-making.