The short-term traffic flow prediction is of great importance for traffic control and guidance. This paper presents an approach using a Sugeno fuzzy inference system whose input space is participated by a Gaussian mixture model and parameters are estimated by the least square estimation method. The proposed approach was evaluated on a benchmark problem of the Mackey-Glass time series and the collected traffic flow data via a comparison made with one of well-known methods. The experimental results indicate the proposed method is effective and competent.