Airline market segments after low cost airlines in Thailand: Passengerclassification using Neural Networks and Logit model with selective learning

Competition in airline business is severe after an introduction of low cost airlines. In Thailand, three low cost airlines occupied one-third of domestic market at the end of 2005. Their growth rate, 47 percent, surpassed the industrial growth rate at the expense of full-service airlines. One million passengers of full-service airlines were lost to low cost airlines in 2005. The competition drives airlines to clarify their market segments. Passenger information is crucial for retargeting and repositioning. In this study, questionnaires were collected from 468 Thai passengers at Chiang Mai International Airport during October to November 2005. Clients of full-service airlines and low cosst airlines shared equally in the allocation of questionnaires. Neural Networks, an alternative technique for airline passenger classification, was benchmarked to a traditional econometric model, Logit. Information from 368 passengers was included into the learning process of models whereas 100 were used for validation. In prediction, Logit model showed little advantage over Neural Networks. However, transmission of only significant variables from Logit model to the learning process of Neural Networks, the selective learning, raised 7 percentage points in accuracy over mere Neural Networks and 2 percentage points over Logit model. Based on the prediction, 64 percent of Thailand’s domestic air passenger transportation could be clearly separated into two dominant markets for full-service airlines and low cost airlines. The remaining 36 percent was still an overlapping market segment. Tourist was a significant group in this overlapping segment. Therefore, capturing tourists’ preference will yield higher advantage in the airline business competition.