Back-propagation neural network combined with a particle swarm optimization algorithm for travel package demand forecasting

Travel agency operation must be capable of judging tourism demands on the market to formulate procurement and sales plans. However, a number of travel agencies, lacking the ability to judge market tourism demand, take substantial risks in their purchasing decisions. Therefore, in this study we used the particle swarm optimization algorithm combined with a back-propagation neural network (PSOBPN) to establish a demand estimation model and we used gray relational analysis to select factors highly correlated to travel demand to use as training and prediction input factors in the prediction model. A comparison of the prediction results with the back-propagation neural network (BPN), multiple regression analysis (MRA), and the travel agency experience forecasting method indicated that the PSOBPN and BPN prediction accuracy were superior to that of MRA and the experience forecasting method adopted by travel agencies. The accuracy of both PSOBPN and BPN were equal. The convergence speed during PSOBPN training was superior to that of BPN. PSOBPN is a superior prediction model when considering both prediction accuracy and model training convergence speed. It can provide decision makers for travel agencies with reliable and highly efficient data analysis.