Application of Prophet-LSTM combined model in prediction of air transportation incidents
To achieve accurate predictions of the air transportation incident per 10 000 flight hours in China,a novel method that combines time series and neural network models was proposed.First,a Prophet model was established using the air transportation incident per 10 000 flight hours data from January 2008 to December 2020.The RStudio software was used to fit the model and obtain the linear part of the air transportation incident per 10 000 flight hours.Secondly,an Long Short-Term Memory(LSTM)neural network model was used to capture the nonlinear part of the air transportation incident per 10 000 flight hours.Lastly,the Prophet-LSTM combination model was established using the reciprocal variance method.The combination model was used to predict the air transportation incident per 10 000 flight hours from January to December 2021,and the predicted results were compared with the actual values.It can be concluded that the three models'responses to the periodic fluctuations and evolution trend characteristics of time series data are generally consistent with the actual situation from the predicted curve chart.All three models can be used to evolve the patterns of air transportation incidents.However,the effectiveness of prediction is measured by the size of three indicators;EMA,EMAP,and ERMS.The smaller the values of the three indicators are,the higher the prediction accuracy of the model is.The results show that the EMA,EMAP,and ERMS of the Prophet-LSTM combination model are 0.097 3,16.128 5％,and 0.128 7.The EMA,EMAP,and ERMS of the Prophet model are 0.123 3,20.046 5％,and 0.150 8.The EMA,EMAP,and ERMS of the LSTM model are 0.098 8,16.309 0％,and 0.132 5,respectively.Compared with the single model,the precision of the Prophet-LSTM combined model is significantly improved,respectively.Compared to the existing ARIMA+BPNN combination model and GM(1,1)+ARIMA+LSTM combination model,the Prophet-LSTM combination model reduces the EMA,EMAP,and ERMS by 0.025 9,10.487 4 percentage points,and 0.014 3,respectively,and 0.012 8,2.059 9 percentage points,and 0.008 6.The results demonstrate that the Prophet-LSTM combination model has higher prediction accuracy and better performance.
safety social engineeringair transportation incidentProphet modelLong Short-Term Memory(LSTM)modelcombination prediction model
万方数据
维普
