Our business goal is try to better allocate human resource in Northeast and Yilan Coast National Scenic Area Due to the tourism industry booming in Taiwan the past five years, the domestic and foreign tourists are increasing. we select ocean park, hiking trail, rock climbing area, diving space and camp site. These five locations are directly under different units. Try to organize the staffs between these five locations. Our data comes from Tourism Bureau Executive Information System, and the data is measure the monthly tourist volume in Northeast and Yilan Coast National Scenic Area from 2005 to 2013.

Because there are trend and seasonality in most series, to choose the proper forecast methods, first we chose the seasonal naïve to be our benchmark. Second, we also use the Holt-Winters exponential smoothing to be our model which allows the level, trend, and seasonality patterns to change over time. The third method we used is multiple linear regression, and we used month to be our dummy. We drew the forecast and actual charts and residual charts, and we compared the residual chart of these three methods. And we can see in this table, we also compared the RMSE in these three methods.

There are some limitation in this forecasting project. First, we can’t find out the difference of tourists volume between weekday and weekend due to lacking of daily data. Second, because there is a big difference of tourist volume between the peak season and low season in Fulong ocean park, it is difficult to forecast the tourist volume accurately. Third, we can’t find out the relation of tourist amount between these five places using the data in hand. Finally, it is difficult to give recommendation about human resource allocation in detail because we don’t have real human resource arrangement data.

And our recommendations for the Longdong South Ocean Park and Caoling Historic Trail system and Fulong Ocean Park, can use Holt-Winters. On the other hand, in the Longdongwan Coast Park) and Longmen Camp Site, it can simply use seasonal naive will cause the best result.