According to the problem that Gap Statistics (GS) method can get rough classification of data sets and can not get the fine classification of data sets, this paper improves GS method by introducing Gap statistic into ISODATA algorithm and proposes IGS model. Empirical research indicates that IGS model can not only realize the fine classification of data but also can estimate the optimal number of classification of the data sets through IGS model whose accuracy is obviously higher than original GS model.