Modelling passengers' show-up prediction for inventory management of airline industry in Malaysia

This research studies the inventory management of an airline industry in Malaysia. In airline industry, no shows are inevitable when handling the passengers on the departure day. Therefore, the overbooking process has been introduced to cope with this issue and aimed to improve the flight optimization. However, the performance of the flights‟ load factor reflects a moderate seats‟ utilization and this allows for further improvement. Further analysis needs to be done by the inventory analysts to improve the flights‟ load factor. Due to high volume of daily ad hoc requests that need to be answered, inventory analysts have a limited time to focus on the analysis. In addition, it is found that there are very limited previous studies found on predicting passenger‟s show up rate in the airline industry and none of the papers found was conducted in Malaysia. Thus, the main objective for this study is to advocate operational methods that could assists the analysts on the show up analysis and their daily inventory decisions. Specifically, the study aims to formulate a logit model to predict the passengers‟ show up probability on the departure day. Extending the usage of the model, this study aims to generate a scorecard and a decision tree as tools to assist the analysts on their daily operation tasks. Using SPSS 22, the show up prediction is modelled by using the logistic regression approach and linear discriminant analysis. The logit model is found to correctly predict 78.5% out of the validation set. Result from the Receiver Operating Characteristic (ROC) curve shows an area under the curve of 0.708 and depicts a satisfactory model. The discriminant function obtained from the linear discriminant analysis resulted in 97.5% correct prediction. A scorecard is developed in extension to the logit‟s results using Microsoft® Office Excel 2010. The Weights of Evidence are calculated using the formula defined and tailed by the calculation of scores for each significant attributes. The sample scorecard is shown in Chapter 4. A decision tree is also developed to support the task, but focusing more on the analysts‟ daily inventory decisions i.e. handling ad hoc requests by phone bookings and emails. From the result, the tree generated a correct prediction of 81.5% in the validation process. The scorecard and the decision tree developed are believed to be helpful to the inventory analysts in strategizing the flight seats in pursue to a better flight optimization. Based on the previous studies done, using scorecard to predict the passengers‟ attendance on flight departure, as demonstrated in this thesis, is a new approach to the industry. Hence, this research aims to advocate the scorecard method and also utilize the decision tree method to help the flight analyst strategies efficiently in their daily inventory management.

Text IPM 2015 3 D.pdf Download (1MB)

Actions (login required)

View Item