Citation count prediction based on Google Scholar profiles and Clarivate’s journal citation reports

What the paper says

Introduction. Citation count prediction (CCP) models are vital for assessing research impact, yet existing approaches suffer from critical limitations. Prior studies often rely on restricted datasets (e.g., journal metrics alone) or fail to account for the multidimensional factors influencing citations, leading to suboptimal accuracy. Method. We propose an accurate CCP regression model for Computer Science and Electrical Engineering disciplines found on twenty three novel features extracted from public data in Google Scholar profiles and the Journal Citation Reports (JCR) annual report by splitting features into four datasets: Author information database (AI DB), journal information database (JI DB), paper information database (PI DB), and finally author & paper & journal information database (APJ DB). Analysis. Our evaluation employed Mean Absolute Error (MAE), Mean Squared Error (MSE), Mean Absolute Percentage Error (MAPE), and the Coefficient of Determination (R²) to assess model performance. Dimensionality reduction techniques like Principal Component Analysis (PCA) and t-Distributed Stochastic Neighbor Embedding (t-SNE) were also applied, and their effect on CCP was assessed. Results. We identified that paper-level features (PI DB) were significantly more predictive than author or journal attributes, resolving a key debate in CCP research. Conclusions. This study enhances CCP research by introducing scalable, publicly available features, demonstrating the superiority of paper-level attributes through empirical evidence, and identifying Nu-SVR as the most effective algorithm for accurate and interpretable citation prediction, supporting researchers, institutions, and policymakers in assessing research impact.