Research on intercity travel mode recognition and network structure characteristics based on complex network and random forest classification

This study aims to address the issue of identifying diverse intercity travel modes in high population density areas and analyzing their network structure characteristics over different time periods. Consequently, a management model combining complex network theory (CNT) with a random forest classification (RFC) algorithm-referred to as the CNT-RFC model-is proposed. The study utilizes publicly available migration data from the National Bureau of Statistics and transportation departments from January 2021 to December 2023. Network structure features are extracted through node degree distributions, edge connections, centrality metrics, and community detection algorithms. Integrating RFC enables precise identification of travel modes and uncovers the spatio-temporal heterogeneity of intercity travel patterns. Experimental results demonstrate that the optimized model outperforms comparative methods on key metrics including accuracy, precision, recall, and F1 score. Specifically, for leisure travel identification, the CNT-RFC model achieves an accuracy of 0.947, precision of 0.928, and F1 score of 0.947, surpassing advanced models such as the Funnel Tabular Transformer and Graph Transformer for Node and Edge Representation Learning. Paired sample t-tests confirm that these improvements are statistically significant (p < 0.05) with a very large effect size (Cohen's d > 3.5). Network structure analysis reveals a decline in the small-world coefficient to 0.58 during holiday periods, an increase in average travel distance to 25 km, and a rise in average adaptation cost to 0.30, indicating significant structural reconfiguration. Sensitivity analysis related to the pandemic further validates the model's robustness, showing only a slight accuracy decline of 0.0101 in 2022, while the centrality of the high-speed rail network decreased by 0.0720, confirming the pandemic's asymmetric impact across transportation network layers. This study provides robust scientific evidence and effective solutions for regional transportation planning and policy formulation, thereby effectively addressing traffic congestion, optimizing travel routes, and improving intercity transportation efficiency.