Scholarlink

Lack of accurate passenger traffic planning proves to be a significant problem for airlines. Challenges like overbooking or underbooking flights, improper crew and aircraft planning, and inventory issues can raise operation expenses, compromise customer satisfaction, and adversely affect profits. Time series forecasting is a crucial tool that enable them to solve the issues of airlines, presenting the prognosis of passenger flow according to historical data. This paper compares four common time series models, ARIMA, SARIMA, Prophet, and Tuned SARIMAX, to identify which model best estimates the future passengers of the airline using data from 2009–2019. The implemented models were executed using various metrics including MAPE (Mean Absolute percentage Error), RMSE (Root Mean Squared Error), and normalized RMSE. SARIMA model produced the best performance with MAPE of 1.05%, RMSE of 76.40, and normalized RMSE of 2.14%, capturing trend and seasonality nicely. Prophet followed with a MAPE of 1.94% and RMSE of 144.15, followed by ARIMA and SARIMAX with MAPEs of 1.76% and 6.13%, respectively. SARIMA was more precise than other models. This analysis highlights the importance of the right model to be chosen to address the special issues of the airline sector. Accurate forecasting can significantly improve operational efficiency, resource utilization, and profitability. The results indicate that SARIMA is the best choice for this case, and future work could be on hybrid models and including more exogenous variables for greater accuracy.

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

Time Series Airline Passenger Traffic Forecasting Using Prophet, ARIMA, SARIMA, and SARIMA with Exogenous Variable

  • 会议论文

  • 发表日期：

    2025年

查看全文
  • Sweeti Sah,
  • Yuvika Pothrapally,
  • Abhishek Nayak Bhukya

摘要

Lack of accurate passenger traffic planning proves to be a significant problem for airlines. Challenges like overbooking or underbooking flights, improper crew and aircraft planning, and inventory issues can raise operation expenses, compromise customer satisfaction, and adversely affect profits. Time series forecasting is a crucial tool that enable them to solve the issues of airlines, presenting the prognosis of passenger flow according to historical data. This paper compares four common time series models, ARIMA, SARIMA, Prophet, and Tuned SARIMAX, to identify which model best estimates the future passengers of the airline using data from 2009–2019. The implemented models were executed using various metrics including MAPE (Mean Absolute percentage Error), RMSE (Root Mean Squared Error), and normalized RMSE. SARIMA model produced the best performance with MAPE of 1.05%, RMSE of 76.40, and normalized RMSE of 2.14%, capturing trend and seasonality nicely. Prophet followed with a MAPE of 1.94% and RMSE of 144.15, followed by ARIMA and SARIMAX with MAPEs of 1.76% and 6.13%, respectively. SARIMA was more precise than other models. This analysis highlights the importance of the right model to be chosen to address the special issues of the airline sector. Accurate forecasting can significantly improve operational efficiency, resource utilization, and profitability. The results indicate that SARIMA is the best choice for this case, and future work could be on hybrid models and including more exogenous variables for greater accuracy.