Abstract
In the aviation industry, maintenance and inventory holding costs of spare parts give managers an opportunity to decrease their operational costs. Therefore, highly accurate demand forecasting is an indispensable entity in spare-parts inventory management. In the literature, traditional demand forecasting methods and measures are insufficient due to the variability of demand quantity and the uncertainty of demand occurrence times. When comparing the demand forecasting methods, MAPE and RMSE measures are usually preferred and these methods often give misleading results when inventory cost minimization is considered. In this paper, a cost-based performance measure and a newly generated joint performance measure, GMEIC, are employed to compare the traditional forecasting methods with the methods generated for non-smooth demand. The methodology applied in this paper consists of data classification, parameter tuning, and ordering decisions based on (Q, R) inventory policy and inventory cost evaluations. In order to measure the performances of the employed forecasting methods for non-smooth demand data, 632 different items were selected from the inventory of Turkish Airlines Technic MRO. It was observed that traditional forecasting methods did not perform better than the forecasting methods developed for non-smooth demand data when the inventory costs were taken into account as the performance measure.