SARIMA模型在航空公司运营安全状态预测中的应用

On the application of SARIMA model to predicting status of airline operational safety

应用季节性差分自回归移动平均模型(SARIMA)对X航空公司动态安全指数进行预测分析,为航空公司运营安全的规划和发展提供决策依据。收集整理X航空公司2005年1—6月的安全事件和运行数据。利用动态安全指数计算方法对数据进行预处理,建立时间序列。应用SPSS软件对动态安全指数的时间序列进行模型拟合,建立SARIMA模型。对所获得的模型进行参数检验,选取最优模型。利用最优模型对2015年7—12月动态安全指数进行预测,并对预测值与实际值进行对比分析。结果表明,SARIMA(1,0,2)(0,1,0)12模型在显著性水平0.05下通过了所有参数检验。各月实际值都落入了拟合值95%的可信区间范围,动态安全指数的实际值与拟合值变化趋势基本一致。2010年之后精度较高,实际值与拟合值具有较好的重合度。△ln Yt拟合值的最大绝对误差为1.976 6(2009年12月),最小绝对误差为0.000 4(2013年9月)。2015年7—12月,动态安全指数的实际值与预测值变化趋势基本一致,但误差较大。SARIMA模型能够较好地短期模拟X航空公司动态安全状况和趋势,预测效果良好。当发生事故、严重事故征候时,序列的实际值会偏离序列原有的结构,预测精度下降。

The present paper is aimed at providing and proposing the SARIMA model for analyzing and forecasting the airline operational safety status quo.For the said purpose,we have managed to analyze the dynamic safety index（DSI） of some airline company known as X and worked out the changing regularities and patterns for the safety performance control on the basis of the civil aviation decision-making basis and the safety plan-making,multiple seasonal autoregressive integrated moving average（SARIMA）model.In proceeding with our study,we have collected a lot of operational data and the statistics of the monthly flying accidents and incidents from Jan.to Jun.2015 in the said X airline with the calculation model of DSI and the data being processed and the time series being set up,whereas the SPSS has been used to fit the said time series.And,then,in the modeling process,the paper has adopted 13 operational indicators of the said X airline as the influential factors,while it has chosen to use the optimal model and the parametric testing results.In addition,we have also adopted the optimal model to predict the monthly DSI in the second half of the year 2015,that is,from July to Dec of the year.The results of our calculation and investigation indicate that the SARIMA（1,0,2）（0,1,0）12can successfully experience and pass through the parametric tests with the confidential index of 0.05,including 95% of the confidential intervals of the values fitted.What is more,the real values and the fitted values prove to be in a nice accord.Moreover,the maximum absolute error of Δln Yt tends to be about 1.976 6（Des.2009）,whereas the minimum absolute error of Δln Ytturns to be merely 0.000 4（Sep.2013）.The same tendency has been properly kept between the real value and the predictive values from July to Dec.in 2015,though there exist considerable errors with the calculation results.Thus,it can be concluded that the SARIMA model can be used to make a short-term forecast for the DSI,nevertheless,there still remain quite a number of uncertain factors that are likely to influence the aviation safety,e.g.the macroeconomic environment,the mergers and the acquisitions,and/or the flight schedule regulations,etc.,which may imply the poor forecast accuracy in the longterm＇s point of view.Additionally,if there took place an accident or an incident in the flying management,it would be valueless or deviated from the original inclination to improve and heighten the airport flight control quality through innovative reform and change,which may in turn greatly run from or dismay the forecasting value of the model.And,perhaps,the outliers＇ examination method can be utilized to improve the prediction accuracy of the model.