查詢結果分析
來源資料
相關文獻
- 傳統計量迴歸、模糊迴歸、GMDH、類神經網路四種方法在預測應用之比較--以國人赴港旅客需求之預測為例
- Neural Network Procedures for Taguchi's Dynamic Problems
- A Fast and Efficient Competitive Learning Design Algorithm Based on Weight Vector Training in Transform Domain
- 專家系統振動訊號圖型判別之研究
- 反傳遞模糊類神經網路於流量推估之應用
- 類神經網路(Neural Networks)的種類及其在影像處理上的應用
- C++Fuzzy類神經網路物件導向發展系統之建立
- 臺灣汽保費率之估計--對數線性費率模式與類神經網路之比較
- 運用類神經網路於股價指數之套利--以日經225指數為例
- 使用類神經網路預估碳化鎢材料放電加工性能
頁籤選單縮合
題 名 | 傳統計量迴歸、模糊迴歸、GMDH、類神經網路四種方法在預測應用之比較--以國人赴港旅客需求之預測為例=The Comparison of Four Kinds of Prediction Methods: Traditional Econometic Regression, Fuzzy Linear Regression, GMDH, and Atrifical Neural Network--An Example of the Demand Forecast of the Taiwanese Tourists to Hong Kong |
---|---|
作 者 | 曹勝雄; 曾國雄; 江勁毅; | 書刊名 | 中國統計學報 |
卷 期 | 34:2 1996.06[民85.06] |
頁 次 | 頁132-161 |
分類號 | 550.19 |
關鍵詞 | 傳統計量迴歸; 模糊迴歸; 類神經網路; Traditional statistical regression; Fuzzy regression; GMDH; Artificial neural network; |
語 文 | 中文(Chinese) |
中文摘要 | 本文目的在比較傳統計量迴歸、模糊線性迴歸、GMDH及類神經網路四種不同的預測方法,並以國人赴港旅客需求人數之預測為例,進行四種預測方法在應用上之比較。模糊線性迴歸模式基本上是一種區間預測模式,而此模式極易受到極端值的影響而使得預測區間變大;GMDH方法則假設自變數與應變數之間成高階多項式關係;類神經網路則是經由梯度下降法修正網路層間的權數值,使得網路計算輸出值與期望輸出值間的誤差逐漸降低。由本例的實證結果發現,傳統計量迴歸、GMDH、類神經網路之預測能力均極佳。若以誤差百分比而言,GMDH在前幾期之變化起伏較大,明顯地較傳統計量迴歸與類神經網路模式為差,但在第六期之後則較其他兩種模式為佳;而傳統計量迴歸與類神經網路模式兩者在整體上差距不大。至於模糊迴歸則因模式輸出為一區間,因此無法與其他三種模式同時比較。 |
英文摘要 | The purpose of this paper is to compare four kinds of prediction methods, including tranditional econometric regression, fuzzy linear regression, GMDH, and artificial neural network. The comparison will be based on the performance of these methods applied to the demand for travelers from Taiwan to Hong Kong. The fuzzy linear regression model is basically an interval prediction model. This method has a disadvantage that the prediction interval can be very wide if some extreme values are present. GMDH presumes that the dependent variable and the independent variables have a high-order relationship. The artificial neural network employs the gradient-descent method to revise the weight values among the network hierarchy. As a result, the error between the output value and the target value will be lowered gradually. It is found from the empirical results of this example that the prediction capability of the traditional statistical regression, GMDH, and the artificial neural network are rather encouraging. For the error expressed in terms of percentage, the upheaval of GMDH in the first few periods is quite prominent, and it is obviously inferior to the traditional statistical regression and the artificial neural network. However, after the sixth period, GMDH performs better than the other two methods. As for the traditional statistical regression and the artificial neural network, they performs about the same as a whole. The fuzzy regression cannot be compared with the other three methods because the output of this methods is an interval. |
本系統中英文摘要資訊取自各篇刊載內容。