Predicting the Economic Impact of Scientific Publications in Biotechnology Using Machine Learning

The economic impact of research papers reveals the diffusion of information and its applicability to other technical fields. This research aims to predict the number of academic paper citations in patents. Papers gathered as a dataset for the study are the outputs of Iran's biotechnology field, indexed in the Scopus database from 2003 to 2024. To conduct the research, 15 indicators have been extracted for these articles in five categories: Journal, Altmetrics, Impact, Open Access, and Collaboration. We performed data processing, exploratory data analysis (EDA), machine learning modeling, and predictions using Python and libraries such as Pandas, NumPy, Matplotlib, Seaborn, and Scikit-Learn. The findings indicated that strong positive correlations are observed between the "Cite Score" and "SJR" indices, reflecting their related nature in evaluating journal impact. The "impact" category shows the strongest positive correlation with "patent information." The "journal" and "Altmetrics" categories show significant correlations, albeit to a lesser extent, indicating their complementary role in predicting economic impacts. Journal category indices, including SNIP, CiteScore, CiteScore percentile, SJR, and SJR percentile, exhibit a range of correlations with Patent citations. Altmetrics indices show a positive correlation with patent citations, which means that articles with higher visibility and engagement have a more significant impact on the patent literature. The results suggest that while machine learning is a powerful tool for predicting economic impact, further model refinement, feature selection, and more advanced techniques are necessary to achieve more accurate predictions. Considering the large gap between scientific papers and applied research in Iran's biotechnology field, is essential for managers and policymakers to identify and remove obstacles to the commercialization of scientific advancements.