Bridging the gap between academic exploration and industrial reality in supercritical fluid extraction: a machine learning and kinetic approach

Abstract of the related article: Supercritical Fluid Extraction (SFE) has evolved from a niche technique to a pivotal platform for the circular bioeconomy. This study presents a critical, data-driven assessment of the SFE landscape, integrating text mining, kinetic modeling, and machine learning (Random Forest) to analyze 1,204 scientific articles (2010–2024) and 3,844 patents. The results reveal a "science-push" phenomenon, a structural shift towards microalgae and waste valorization, and a strategic divergence between academic exploration and industrial intellectual property protection.

Description of the data files: This dataset contains the processed data and scripts used to generate the results and figures presented in the article.
1) Artigos_SFE_FINAL_COMPLETO.xlsx: The harmonized bibliographic dataset of 1,204 scientific articles. It includes standardized columns for Matrix Categories (e.g., Marine Sources, Agro-Waste) and Target Compounds, cleaned via text mining algorithms.
2) Tabela_Cinetica_K.xlsx: The output of the kinetic modeling analysis, containing the calculated growth rate constants (k), Standard Errors, and R2 values for all studied matrix and target categories (2010–2024).
3) Scripts_R_Analysis.R: The complete R programming script used for:
a) Data harmonization and cleaning.
b) Kinetic modeling (non-linear regression).
c) Multivariate analysis (Random Forest and Negative Binomial Regression).
d) Generation of figures and patent landscape plots.

Methodology: Data were retrieved from Scopus, Web of Science, and PubMed (scientific articles) and Lens.org (patents). Text mining and statistical analyses were performed using R (v. 4.4.1) with packages dplyr, stringr, minpack.lm, and randomForest.