Learning-Based Optimization Algorithms for Routing Problems: Bibliometric Analysis and Literature Review

Learning-based optimization (LBO) algorithms have exhibited considerable advantages in solving routing problems. In this study, 831 papers published over two decades (2003–2024) are retrieved from the Web of Science database. This work aims to build extensive knowledge maps of LBO algorithms for routing problems by using a scientometric review of new developments and global trends. Prolific journals, conferences, authors, and institutions are discussed in the statistical analysis. The overall trend of LBO algorithms for routing problems is growing, and it is dominated by China and the USA. Collaboration network, co-citation analysis, and emerging trend analysis are developed to identify major disciplines of LBO algorithms for routing problems. Different emphases on the research field in operations research and computer science communities are identified respectively. Studies on LBO algorithms are reviewed from the perspectives of supervised learning (SL), unsupervised learning (UL), and reinforcement learning (RL). The major characteristics and limitations of LBO algorithms in each category are discussed. Dependence on sample labels and cluster numbers restricts the practical application of SL and UL to routing problems. Meanwhile, RL approaches, such as the deep Q-network, which exhibit fast convergence and computational efficiency, have elicited widespread attention in recent years. This study provides meaningful guidance and future direction to designing LBO algorithms for routing problems.