Modelling car ownership in Turkey using neural networks

The present study developed an artificial neural network (ANN) approach for the forecasting of car ownership in Turkey using socio-economic and transport-related indicators. Due to the lack of disaggregate data, the model was based on aggregate data. The actual forecast was obtained using a feed-forward neural network, trained with a back-propagation algorithm. In order to investigate the influence of socio-economic indicators on car ownership, the ANN was analysed based on per capita GDP, petrol prices, car prices and road lengths along with historical car ownership available from 1971 to 1996. A comparison between model predictions and car ownership data in the test period was used for model validation. The projections were made with scenarios which were developed in order to make forecasts up to 2020. The results show that the ANN application model was more successful and reliable than the other classical models in terms of nonlinear reflection ability.