The Study on Visitor's Forecasting Model of Penghu National Scenic Area

• Main Authors: Jwo-Han Tsai, 蔡卓翰
• Other Authors: Jau-Shin Hon
• Format: Others
• Language: zh-TW
• Published: 2010
• Online Access: http://ndltd.ncl.edu.tw/handle/98621472063438879753

碩士 === 東海大學 === 工業工程與經營資訊學系 === 98 === The purpose of setting up National Scenic Area is to provide visitors with natural tourism resources in order to enhance recreational functions. In recent years the influx of large number of tourists has caused Penghu's tourism resources to disappear gradually. Moreover, Penghu tourists showed a strong seasonal variation so that it is hard to coordinate the peripheral industries to support the needs. Therefore, to establish a predictive model of tourists in response to the external environment variation and the recreational resources supply and distribution is absolutely important for the sound management of National Scenic Area. the scholars’ study showed that if using the quantitative method to predict it will draw better result. In Addition, some studies make comparison between multiple measures while lack of thoroughness in the realm of prediction, therefore there may be missing of important information. To cope with the above deficiency, this study uses a combination of Time Series and Back- Propagation Neural Network (ARIMA-BPN) prediction model; expect to bring both linear and nonlinear relationship of data into account to do the quantitative prediction of the tourists study. First, use Time Series ARIMA to establishment the prediction model, then input the error of predicted value to Back-Propagation Neural Network (BPN). Besides, we input ten factors as the reference indexes. Then use Regression Analysis (Best Subset Regression) for variable screening and initiate prediction learning in neural network. This process can revise the predicted value generated by Time Series model to get better result. Empirical results show that, the prediction performance of ARIMA-BPN after variable screening is the best, simple ARIMA model is the second, and the prediction performance of ARIMA-BPN without variable screening is the worst. Therefore, The ARIMA-BPN combination model which proposed by this study confirms that if bringing the linear and nonlinear relationships of data into account simultaneously, then using Regression Analysis to screen variables, can indeed enhance the forecast performance ; but if we don’t proceed with the variable screening, it may result in relatively poor prediction accuracy. We expect the study findings will provide information for the Penghu and the other National Scenic Areas as a reference of management and planning in the future.