近年來,學者們利用支援向量機的基本想法於多變量模糊(非)線性迴歸,獲得求解的計算效率。然而求解模糊支援向量迴歸模式仍然複雜,並且當參數皆為模糊數(例如: )時仍無法求解,因此我們採用Carlsson & Fuller (2001) 提出的模糊數可能性平均數當作限制式,建構了更簡易求解的模糊支援向量迴歸模式,依照參數定義為模糊數與否一共有了六種不同的模式。 透過在資料分析的圖中,我們可以發現模式5 求解出的預測結果與原始資料配置的程度相當高,以誤差均方根(Root Mean Square Error ; RMSE)越小,模型預測精確度越高的特性來衡量預測模型優劣,RMSE達到了1.3134。於是我們根據 Carlsson & Fuller (2001) 提出的模糊數可能性平均數這個概念,所建構的模糊支援向量迴歸模式是一個可行的方法,並且應用於預測上相當精準。 In recent years,introduce the use of SVM for multivariate fuzzy linear and nonlinear regression models with efficiency solutions. However, fuzzy support vector regression model is still complicated to slove the parameters which are all fuzzy numbers. In order to cope with this problem, we adopt the fuzzy possibilistic mean method proposed by Carlsson & Fuller (2001)which is more easily to slove fuzzy support vector regression model. According to parameters are fuzzy numbers or not, there are six kinds of models. Fnally, in data analysis, we can find forecasting vales in our proposed models are fitting very well using RMSE. It is obviously that our proposed fuzzy support vector regression model can be applied to forecast with better forecasting performance