Fuzzy time series approaches have being increasingly attracted researchers' attentions. The procedures on fuzzy time series actually consist of three stages; fuzzification, determination of fuzzy relations and defuzzification. Researches are generally concentrated on these stages and about improving them. In this study, we propose a new approach, which combines several techniques. In this approach, Gustafson-Kessel, which is a fuzzy clustering technique, is being used to fuzzification of time series. The proposed method is compared with the approaches in literature.