Utilization of Genetic Programming to Establish Demand Forecast in Taiwan International Flights

Accurately prediction is the most important way to cost down for airlines. The study was focused on build up forecast model of five Taiwan international flights included Bali, Bang Kong, Ho Chi Minh City, Kuala Lumpur, and Singapore. Genetic programming was adopted to establish simulation models, and Mean Absolute Percent Error (MAPE) also was used to evaluate the performance of those models. The ten years of historical passenger´s data was collected and analysis, and finally the demand forecast of five flights in 2010 would be conducted. The validations MAPE of models were lower than 10% expect Bali flights. Based on experience of traditional statistic method included linear regression and time series, the ability of Genetic programming models were excellent. The forecast error of Bali flights were 11% and it may be caused by a series accident. On the basis of above results, Genetic programming could be the feasible approach for prediction of five flights in Taiwan. In addition, the passengers to Singapore would substantially increase in 2010-2011, and the issue is worthy to further study for airlines and government.