Mapping the landscape of energy markets research: A bibliometric analysis and predictive assessment using machine learning

Abstract

This study examines the evolving dynamics of research on the energy market that focuses on understanding its interactions and interdependencies with other markets. Using published articles from 2002 to 2022 indexed by the Web of Science, we employ bibliometric methods, complex network measurements, and machine learning algorithms to analyze trends and predict academic success or interest. Our bibliometric analysis highlights the growing emphasis on new topics, such as clean energy, over traditional energy topics like crude oil and volatility spillovers. In a horse-race setup, we use supervised regression techniques to predict the paper’s academic success, measured in terms of the average number of citations over the years. We use meta-information from the paper, including keywords, as predictive attributes. The Random Forest achieves the best out-of-sample performance. We complement this analysis by using Shapley Additive Explanations to assess the contribution of each attribute to the overall prediction, thus allowing model interpretability. We find non-linear relationships between some numeric attributes, such as the number of keywords in the paper, and the target variable, highlighting the flexibility of non-linear methods compared to linear ones. Our findings offer valuable insights into emerging research trends and provide educators, policymakers, and finance professionals with critical information to navigate the evolving landscape of energy market research.