In order to make support vector machine（ SVM） better accomplish parameters optimization and obtain forecasting fumction of time series change space

the technical advantages of algorithmic role division mechanism are strengthened by improving the search mode of different bee colonies in artificial bee colony（ ABC） community

and introduce Theory of Fuzzy Information Granulation（ TFIG） to further optimize the learning efficiency

regression precision and generalization ability of prediction framework. Considering that both the prediction precision and generalization performance are better than the existing classical methods when the proposed method is applied to the modeling of the SSE index.The proposed method has good time series prediction function

some inspiration and reference significance for the study of time series forecasting

satisfactory approximation and fuzziness problem in big data background.