The network traffic data is chaotic, and in traditional chaotic forecasting algorism, the Euclid distance is used to measure correlation between phase points in phase space. Because of limitation of Euclid distance, in high dimension phase space, the forecasting accuracy of the traditional chaotic forecasting algorism decreases rapidly. The included angle cosine instead of Euclid distance is used as the standard to judge correlation between phase points. And regarding phase points as vectors and taking vectors module and angles as optimization objectives, identifies forecasting parameters. Applies the above method to forecast the network traffic data and the results indicate that the forecasting accuracy increases remarkably in high dimension phase space.