基于 Prophet-ARIMA 模型的民航周转量-《计算机技术与发展》

文章信息/Info

Title:: Application of Prophet-ARIMA Combined Model in Forecast of Civil Aviation Turnover

作者:: 刘铭基¹ ; 田雅楠¹ ; 张亮¹ ; 金博²; 1. 东北财经大学国际商学院,辽宁大连 116025;
2. 大连理工大学创新创业学院,辽宁大连 116024

Author(s):: LIU Ming-ji1 ; TIAN Ya-nan1 ; ZHANG Liang1 ; JIN Bo2; 1. School of International Business,Dongbei University of Finance and Economics,Dalian 116025,China;
2. School of Innovation and Entrepreneurship of DUT,Dalian 116024,China

Keywords:: Prophet model; NeuralProphet model; turnover forecasting; machine learning; combined forecasting; time series forecasting

摘要:: 周转量作为计算运输成本、客货运收入、劳动生产率、客货运平均行程和运输密度等指标的依据,能比较全面和确切地反映运输的成果以及运输生产产品的数量,其预测对民航的科学化发展有重要意义。与民航业的快速发展和民航市场的不断扩大相比,目前民航的预测模型种类较少。为探索一种更为有效的方法来提高民航周转量预测准确率,较为新颖的 Prophet 模型和 NeuralProphet 模型被引入到对民航货物周转量、民航货邮周转量、民航旅客周转量和民航总周转量的预测中。在与单个模型对比中,在精确度上 Prophet 模型和 NeuralProphet 模型相较于传统的三次指数平滑法以及 ARIMA模型预测结果更优。利用权值法创建的 Prophet-ARIMA 组合模型使预测结果更为精准,并被发现在讨论的所有模型中该模型表现最佳,这为民航预测提供了一种新思路。

Abstract:: As the basis for calculation of transport costs,passenger and freight income,labor productivity,average passenger and freighttravel and transportation density, turnover can reflect the results? of transportation comprehensively and accurately and the number ofproducts produced in transportation,and its prediction is of great significance to the scientific development of civil aviation. Comparedwith the rapid development of civil aviation industry and the continuous expansion of civil aviation market, there are few forecastingmodels for civil aviation at present. To explore a more effective way to improve civil aviation prediction,we introduce novel Prophetmodel and NeuralProphet model pairs to predict civil aviation cargo turnover, cargo and mail turnover, passenger turnover and totalturnover of civil aviation. In comparison with a single model,it is found that the Prophet model and the NeuralProphet model predictbetter results compared to the traditional model cubic exponential smoothing method and ARIMA model. In order to further optimize themodel and make the prediction results more accurate,a weighting method is used to create the Prophet-ARIMA composite model. Theresult is the Prophet-ARIMA model has the best performance compared with other models,which provides a new idea for civil aviationforecast.