Journal of Systems & Management ›› 2020, Vol. 29 ›› Issue (6): 1169-1176.DOI: 10.3969/j.issn.1005-2542.2020.06.014

Previous Articles     Next Articles

Forecast of Travel Index of Air Passengers Based on LSTM-SVR Model

XIONG Honglin,JI He, FAN Chongjun,YANG Mengda   

  1. Business School,University of Shanghai for Science and Technology,Shanghai 200093,China
  • Online:2020-11-29 Published:2021-01-19

基于LSTM-SVR模型的航空旅客出行指数预测

熊红林,冀和,樊重俊,杨梦达   

  1. 上海理工大学 管理学院,上海 200093
  • 通讯作者: 樊重俊(1963-),男,教授。
  • 作者简介:熊红林(1984-),男,博士生。研究方向为信息系统工程、机器学习和智慧机场。
  • 基金资助:
    国家自然科学基金资助项目(71774111);上海市教育委员会科研创新重点基金项目(14ZZ131)

Abstract: The situation of air passenger travel is of great significance to the construction and operation of civil aviation airports. The travel index of air passengers is defined, and the machine learning method is used to predict the travel index of air passengers. To overcome the insufficient prediction accuracy of a single model, a long short-term memory network (LSTM) and support vector regression (SVR) integrated forecasting model is proposed, and clustering analysis is performed on the forecast result set. The empirical progressive analysis based on airport passengers in Shanghai Airport verifies the feasibility and validity of the forecasting model proposed. The experimental results show that the forecasting model proposed has a higher accuracy than the traditional single forecasting model. At the same time, the forecasting model proposed also has obvious advantages compared with other integrated forecasting models. In addition, based on the K-means algorithm, the travel index of air passengers is clustered and a row rating is given, which provides certain value decision support for airport operation management and travel choice of passengers.

Key words: travel index of air passengers, machine learning, long short-term memory (LSTM), support vector regression (SVR), K-means

摘要: 航空旅客出行的情况对民用航空机场建设与运营具有重大意义,定义了一种航空旅客出行指数,运用机器学习方法对航空旅客出行指数进行预测,克服了单一预测模型精度的不足,提出一种将长短期记忆网络(LSTM)与支持向量回归(SVR)相结合的航空旅客出行指数组合预测模型,并对预测结果集进行聚类分析。以上海机场航空旅客数据为实证,验证了LSTM-SVR组合预测模型可行性与有效性,实验结果显示:LSTM-SVR组合预测模型较传统单一预测模型具有更高的精度;同时,LSTM-SVR组合预测模型与其他组合预测模型相比也有较明显优势。此外,基于K-均值算法对航空旅客出行指数进行聚类分析并给出评级,此举为机场运营管理及旅客出行提供一定的决策支持。

关键词: 航空旅客出行指数, 机器学习, 长短期记忆网络, 支持向量回归, K-均值聚类

CLC Number: