The aim of this research is to probe the best ARIMA model for the daily close prices and return rates of the Foxconn stock, and to forecast the accuracy of prediction. We applied Extended Sample Autocorrelation Function to determine the parameters of ARIMA model. The data of this research came from the Foxconn stock price during January 1, 1999-April 30, 2006. We employed the goodness of fit and residual error test to selects the best model. Finally, we performed the forecast and the analysis of ARIMA model. Our results showed that the best model for daily close prices is ARIMA (0, 1, 4). Moreover, we found that the best model for return rates is ARIMA (0, 0, 4). Although the forecast has a little difference between predicted values and the actual values for both models, all fall into 95% confidence interval.