The Study Of Tourism Demand Forecasting Analysis Based On Anfis

With the rapid development of society and world economy, more and more people come in to pursue high-quality pattern of living after meeting their basic standard lifestyle. High-speed developments of every country inevitably bring about the booming of world tourism industry. As one of the so much important destination countries, the whole china gains a great quantity of benefits, not only economy interest but also the interests of the public and society. Then we come into recognizing that the study on forecasting analysis of tourism industry is so important to guarantee efficient sustainable tourism prospect.A variety of forecasting analysis techniques have been developed for the deeper study of prediction by scholars. Unfortunately, all the techniques we used before are mainly linear equations. We could not approach the really world by using traditional techniques onto the non-linear relations of the world. ANN(Artificial Neural Networks)is a widely received technique, but the poor ability of accurately forecasting and many drawbacks on analyzing the influence that factors'impacts on dependent variables limit its application.Three objectives we want to achieve here by introducing Adaptive Network Fuzzy Inference System (ANFIS) into forecasting analysis on tourism industry are: Firstly, to establish a high accuracy model to provide basis for decision making on tourism planning and project management; Secondly, to reduce the dependence of employed model on data to cut down the costs. So, the model we developed here can been employed by different tourism sectors; Thirdly, to enrich the existing forecasting analysis techniques so that we can broaden the selectivity of tourism researchers.By comparing and summarizing the forecasting analysis techniques commonly used before, the paper finds the necessary study on ANFIS. In the case study chapter, the paper selects Japanese monthly Demand for China (2000.01-2008.12, 108 months in total) as dependent variable which characterized by Number of Inbound Japanese Tourist (NIT) and Expenditures of Inbound Tourist(EIT). Variables are month (Mon), CPIs of Japan/China (CPIJ/CPIC), rate of CPIJ to CPIC ( SPR), parity price of RMB to JPY (ER), annual population of Japan (Pop), annual private final consumption expenditure of Japan (PFCE), annual per-capita private final consumption expenditure of Japan (PFCEPC). Data splitting technique is employed to random separate the data collected into two parts, training data and testing data. The first part data is employed to train models to get parameters and the second part data should be used to test the accuracy. At last, the paper employs Sensitivity Analysis Technique to obtain sensitivities.Based on computing by MATLAB, the three objectives we formerly proposed are all met as what we anticipated before. ANFIS possesses its unique advantages compared to traditional techniques: (1) the best accuracy. Measured in PE, R, MAD, MAPE and RMSPE, ANFIS generates the smallest error. (2) Lowest dependence on data and variables. ANFIS needs two factors, Mon and PFCE, for NIT forecasting and three factors, Mon, CPIC and CPIJ for forecasting EIT. (3) ANFIS shows us the right relationship between variables and dependent variables and obtains cures of sensitivity and surface of sensitivity which could not obtained by other models.Some shortages are summarized at the end of the paper and a lot of new ideas are brought forward for deeper study.