Characteristics of a Highly Cited Article: A Machine Learning Perspective

Machine learning (ML) is a fast-growing topic that enables the extraction of patterns from varying types of datasets, ranging from medical data to financial data. However, the application of the ML methodology to understand the key characteristics of highly cited research articles has not been thoroughly investigated, despite the potential practical guidance that ML can provide for researchers during the publication process. To address this research gap, an ML algorithm known as principal component (PC) analysis is used to detect patterns in highly and lowly cited papers. In this paper, eight features (number of citations, number of views, number of characters with no spaces, number of figures, number of tables, number of equations, number of authors, and title length) are extracted from highly and lowly cited papers, leading to eight PCs (PC1–PC8). PC1 shows that the numbers of citations are positively correlated with the character count and negatively correlated with the title length. PC2 shows that the number of tables is positively correlated with the title length. PC3 shows that the number of figures is positively correlated with the number of tables. PC4–PC8 rank the importance of individual features in the descending order: number of equations, number of characters with no spaces, number of figures, number of views, and then the number of authors. The results of the ML analysis provide interesting and valuable tips for researchers, students, and all academic and non-academic writers who are seeking to improve their citation rates.