Application of Modified Grey Prediction Model to Discuss Number of Tourists in Taroko

： Mei-Lien Kan and Yuan-Bing Lee
： Average daily sales， GM(1,1)， Rolling checking， Optimal grey prediction， Growth trend

： The paper brings together the average daily sales ofa Farmers’ Association supermarket from May, 2010 toDecember, 2010. First, we examine whether the datameet the conditions of modeling, and we determine theaverage amount of daily sales data are in line with theGM(1,1) modeling of the capacity district. Also, it meetsthe conditions of the modeling prediction accuracyhigher than 90%. Due to the traditional grey predictionGM(1,1) model only needs 4 data, but residualmodification grey prediction GM(1,1) mode needs atleast five data. Hence, the GM(1,1) model uses the rawdata from May, 2010 to November, 2010 to predictvalues in December, 2010 to do error analysis. Also, weapply the GM(1,1) rolling test, and use the data of 5groups, 6 groups and 7 groups to calculate six kinds ofGM(1,1) prediction model, and we find the raw datawith minimum error as the grey prediction GM (1,1)model. After determining the optimal grey predictionperiod, we apply six grey prediction methods and theoptimal grey prediction period to predict the values ofDecember, 2010 to do error analysis. We also useaverage error value as a criteria to select the bestprediction model, and by using the top four of the bestprediction model, we are able to predict next threemonths’ growth trend.

Journal of Grey System(Volume 16)
： 139~148
： 2013