This study aimed at the development of the prediction model. The objective is to construct a Grey-GARCH model for the stock data from January 2, 1992 to December 31, 1996. Based upon 1265 samples from the time interval mentioned above, this study will analyze a brand new concept that has never been thought. Generally speaking, the higher of the sample size for GARCH, the better for the description of variation. In other word, the feature for Grey of time-series analysis to investment is effective. According to properties of Deng’s(1982) Grey System Theory, this study will use a few amount of data to establish the forecasting parameters. Each of the values for the related parameters will be updated through on-time process for the short-term earnings. This study will use the Grey prediction of GM(1,1) to deduce GARCH model or namely it as Grey-GARCH model. Furthermore, this study will examine a practical Grey-GARCH prediction model to display an improvement for the better accuracy than the traditional GARCH model. The major contribution of this study may increase the profit of investment.