Forecasting Of Tourist Demand Based On BP Neural Network

With the continuous and rapid development of China's economy and the improvement of people's income, as well as the level of organization, methods of transport, and the facilities available at destination points, China's tourist market is expanding steadily. It will more and more obviously promote China's economy.At present commonly used mathematical models in tourism are based on statistics approaches such as time-series model and regression model(includes linear and nonlinear model),but ANN (Artificial Neural Networks), which has been applied extensively in many fields, is seldom used. Tourism market is restricted by many factors which take on anfractuous connections, and some factors are even instable, thus it is difficult to have enjoyable results with traditional techniques. So it's necessary to study some new techniques for nonlinear problems.ANN is an information processing system based on imitating cerebra neural network structure. which is of many powerful functions such as nonlinear approach, cosmically parallel disposal, self-study, self-organization, fault-tolerating etc. So it is proved to be an effective technique for classification and forecasting.ANN is better than traditional statistics approaches in forecasting, because it can precisely identify the correlations between training samples. Furthermore, ANN is far better than normal traditional statistics approaches while training samples are not rich and random errors is also included in the samples. The statistics time period of tourism arrivals data is generally short, and tourism market is disturbed by many unpredictable factors, ANN is thus a more superior model. At present, BP (back propagation neural network) is one of the most widely applied artificial neural networks. Through analyses, the author holds that it is reliable to apply BP in the forecasting of nonlinear problems such as tourism demand forecasting.In this thesis, based on ANN theory, the author probes into the selections for tourism demand forecasting index, ANN forecasting model, and the whole modeling process.A tourism demand forecasting system based on BP is programmed with C language and constructed in the operation system of Win 98 and some application program modules are called by means of files.Yunnan's tourism demand is forecasted by the application of this system, and the simulating results is satisfactory.In the forecasting of Yunnan's international tourism demand, the model with structure of 2-7-2, 5-8-1 is respectively adopted for the forecasting of overseas tourism arrivals and receipts (per person), thus the overseas tourism arrivals and receipts in 2005-2008 is forecasted. In order to study and compare the feasibility of the system, the author establishes Quadratic and Compound model with the same training sample. The simulation results of different techniques are evaluated with MAPE(Mean Absolute Percentage Error) and R(Pearson Correlation) .The results indicate that the simulating accuracy of the system is better than any other techniques' mentioned.In the forecasting of Yunnan's international tourism demand, In order to study and compare the feasibilities of different models, the author also establishes Cubic regression models (GDP- overseas tourists, GDP- foreigners and GDP-earnings in foreign exchange) with the same training sample. The results indicate that the simulating accuracy of the system is more or less better than any other models' above mentioned.