Efficiency prediction of airports in Turkey: Utilization of data envelopment analysis and artificial neural network

In this study firstly efficiency of airports was found by using Data Envelopment Analysis (DEA). After that an Artificial Neural Network (ANN) model was created to predict efficiency situation for current or new airports by using data which used in DEA and efficiency situation according to DEA. This study takes into consideration 41 Turkish airports. Number of check-in counters, number of baggage conveyors, number of gates (bridges included), number of runways, terminal size, number of employees and autopark capacity are set as inputs, while total freight traffic, number of passengers and number of commercial air traffic movement are set as outputs. The results of DEA show that 19 airports work efficiently in Turkey. After DEA, ANN model development, which can predict an efficiency situation was aimed. For this purpose variables which used in DEA as input and outputs, used in ANN in input layer and efficiency situation of an airport, which found by using DEA used in ANN in output layer. In the training process of ANN was run for different numbers of hidden layers from 1 to 2 and for different numbers of neurons in each hidden layers from 1 to 6. The ANN structure which has an lowest mean square error was selected as a best structure. Best results was obtained from 2 hidden layer structure which has 5 neurons in the first layer and 6 neurons in the second layer. This structure tested by test group which has 6 different airport datas. Model correctly classified efficiency situation of all airports which were in the test group. This model can be used for predicting efficiency situation of airports. Decision makers can evaluate different scenarios or their own performance by using this structure.