Predicting the impact of academic articles on marketing research: Using machine learning to predict highly cited marketing articles

Abstract

The citation count of an academic article is of great importance to researchers and readers. Due to the large increase in the publication of academic articles every year, it may be difficult to recognize the articles which are important to the field. This thesis collected data from Scopus with the purpose to analyze how paper, journal, and author related variables performed as drivers of article impact in the marketing field, and how well they could predict highly cited articles five years ahead in time. Social network analysis was used to find centrality metrics, and citation count one year after publication was included as the only time dependent variable. Our results found that citations after one year is a strong driver and predictor for future citations after five years. The analysis of the co-authorship network showed that closeness centrality and betweenness centrality are drivers of future citations in the marketing field, indicating that being close to the core of the network and having brokerage power is important in the field. With the use of machine learning methods, we found that a combination of paper, journal, and author related drivers perform better at predicting highly cited articles after five years, compared to using only one type of driver.