eduzhai > Applied Sciences > Transportation >

Airline Performance Prediction

  • Save

... pages left unread,continue reading

Document pages: 41 pages

Abstract: Over the last decade the airline industry has experienced extreme financial adversity. Whilst external low probability but high impact factors have been partially to blame some airlines have done worse than others. Predicting differences in airlines’ financial performance is of interest for various stakeholders. A performance, distress or bankruptcy prediction model can be used by financial institutions to assess the financial health of a company in order to calculate the likelihood of recovering a loan or an investment. Such models can also be used as an early warning system in order to initiate change or proactive turnaround. I use a neural network approach to predict airline performance. The industry is international airlines, the data used is non-financial, and input parameters were selected based on significant differences of the means between the two performance states. Focusing on performance opposed to distress, the concern is primarily with prediction accuracy and stability when using future data. Hence, the robustness of the model is based on its ability to detect subtle variations in airlines’ operating characteristics that can cause shift from loss to profit or vice versa. Using multi-year data to construct prediction models is still relatively underexplored. A multi-year data model captures better the year to year changes causing fluctuations in airline profitability. The results show that a multi-year dataset for performance prediction gives more robust and stable prediction accuracy on hold-out data two and three years into the future, than a single year model. The multi-year network demonstrated a fairly high prediction accuracy of 91 percent overall and relatively stable prediction performance in year 2 and year 3, especially for type 2 error. Compared to other usual benchmark models such as Altman (1968) the performance of the model was superior in year 3 and matching Beaver’s (1966) results in year 3. Compared to airline models it had superior performance to Gritta et. al., (2000) neural network model, one year prior. The same applied to the non-financial LRA model specified by Gudmundsson (1998) for U.S. new-entrant airlines, which was outperformed in all three years. The good generalization trait of the multi-year model shows that a performance prediction model for international airlines is attainable and perfectly comparable to single country, single industry financial based models reported so far.

Please select stars to rate!


0 comments Sign in to leave a comment.

    Data loading, please wait...