Research Landscapes, Hotspots and Emerging Trends of Recommender System (2000-2024): A Scientometric Analysis

This paper presents a scientometric analysis of the research evolution of recommender systems spanning two decades, from 2000 to 2024. The analysis delves into the evolving landscapes, focal points of research, and novel trends within this domain, using a dataset curated from the Web of Science. Beyond traditional scientific output assessments, advanced scientometric tools were employed to dissect the intellectual structure of recommender systems research. The findings reveal a substantial increase in the volume of research papers over the past twenty years, with a corpus of 8,826 papers originating from 118 countries/territories. China, The United States, Australia, England, and Spain emerge as the top five most prolific contributors. The research domain encompasses 4,904 institutions, with the Chinese Academy of Sciences, Tsinghua University, University of Technology Sydney, University of Science and Technology of China and University of Chinese Academy of Sciences being the most influential. Furthermore, keywords such as Knowledge Graph, Deep Learning, Feature Extraction, Graph Neural Network, Data Models, Causal Inference and Singular Value Decomposition exhibit significant citation bursts, signifying the cutting-edge directions in recommender systems research. The insights provided in this paper aim to bolster ongoing research endeavors in recommender systems, serving as a beacon for future scholarly exploration.