It is well known that for GM(1, 1) forecasting, it needs only less data quantity while for GM(1, N) forecasting, the more is the data quantity, the more will be the forecasting precision. In the paper, these two models are connected reasonably by self-correlation analysis. It is a meaningful improvement for reducing Errors to achieve GM(1, N)'s forecasting precision but with GM(1, 1)'s data requirement.