Fuzzy c-mean algorithm is sensitive to the starting value and its restraining result is easy to fall into the partial minimum. Thus, two-stage fuzzy c-mean cluster algorithm is proposed. First through the appropriate similar measure (satisfies similarity and nearness) to estimate the classified number, the selection initial cluster center, carries on the cluster again through the fuzzy c-mean algorithm, finally uses the gray Logistic model to the obtained cluster center to carry on the forecast. Because the cluster center has the statistical characteristic, preferably overcome random error bet...