Scholarlink

<p>Airports face increasing pressure to balance passenger experience and operational efficiency as traffic grows, which motivates application-oriented prediction and allocation models for terminal operations. Using monthly indicators from the Civil Aviation Administration of China (2005–2020) across four airports, this study applies a Genetic Algorithm-optimized Extreme Learning Machine (GA-ELM) to passenger-flow prediction and couples the prediction output with a gate-allocation model that minimizes peak waiting-area density. The genetic algorithm is used to tune ELM input weights and hidden-layer biases, reducing the best observed mean squared error from 0.0254 to 0.0168 across the tested parameter configurations. To avoid overstating methodological novelty, the contribution is positioned as an integrated airport-operations application of established GA-ELM and gate-allocation techniques rather than as a new learning algorithm. The allocation experiment reduces the maximum density from 0.73 to 0.35 people/m2 and decreases gate-resource utilization variance by 51.6% under the stated planning assumptions. The model is therefore best interpreted as a decision-support framework for scenario-based gate planning; its operational deployment requires further benchmarking, chronological validation, and robustness testing under disruptions, heterogeneous passenger behavior, and changing gate conditions.</p>

错误：搜索内容不能为空，请输入英文关键词
错误：关键词超出字数限制，请精简
高级检索

Passenger density prediction and gate allocation via GA-optimized ELM

  • 期刊论文

  • 出版模式：

    开放获取

  • 发表日期：

    2026年8月1日

查看全文
  • Lin Yang,
  • Yuefeng Zheng

摘要

Airports face increasing pressure to balance passenger experience and operational efficiency as traffic grows, which motivates application-oriented prediction and allocation models for terminal operations. Using monthly indicators from the Civil Aviation Administration of China (2005–2020) across four airports, this study applies a Genetic Algorithm-optimized Extreme Learning Machine (GA-ELM) to passenger-flow prediction and couples the prediction output with a gate-allocation model that minimizes peak waiting-area density. The genetic algorithm is used to tune ELM input weights and hidden-layer biases, reducing the best observed mean squared error from 0.0254 to 0.0168 across the tested parameter configurations. To avoid overstating methodological novelty, the contribution is positioned as an integrated airport-operations application of established GA-ELM and gate-allocation techniques rather than as a new learning algorithm. The allocation experiment reduces the maximum density from 0.73 to 0.35 people/m2 and decreases gate-resource utilization variance by 51.6% under the stated planning assumptions. The model is therefore best interpreted as a decision-support framework for scenario-based gate planning; its operational deployment requires further benchmarking, chronological validation, and robustness testing under disruptions, heterogeneous passenger behavior, and changing gate conditions.