近年來隨著不動產估價之應用愈來愈廣泛,凡是經濟愈活絡的地區,不動產估價愈有其必要性,因此,如何評估不動產價值為相當重要的議題。本研究主要分為兩階段,首先將主要影響不動產價格之因素,透過適應共振理論從中學習各影響不動產價格之因素的內在分類原則,且藉由警戒值之調整產生不同分類的結果,接著結合類神經網路建立一套可評估不動產合理價格之估價模式。 本研究實證結果如下: 一、經第一階段適應共振理論網路分群所推估之不動產價格準確性優於未透過第一階分群之不動產價格推估結果。 二、影響不動產價格變數透過適應共振理論網路分群後,分群結果能夠作為一項影響不動產之替代變數。 三、透過適應共振理論網路分群且警戒值參數設定為0.5時,載入第二階段倒傳遞類神經網路推估不動產價格,其效果最佳。 四、於實證結果可發現,倒傳遞類神經網路之輸入層,刪除異常點樣本之效果較佳,故於輸入變數時,應該刪除異常點樣本。
Real estate appraisement in recent years as more and more applications, all the more active areas of the economy, real estate Appraisement of the more necessary, therefore, how to assess the real value of an important issue. The study is divided into two stages, first the main factors influencing real estate prices, through the adaptive resonance theory to gain knowledge of factors influencing real estate prices in the inner principles of classification, and the vigilance test adjustments by different categories of results, then using neural networks to establish a reasonable price of Real estate Appraisement model. The empirical results are summarized as follows: 1、The first stage of through the adaptive resonance theory to clustering accuracy of the appraisement of the real estate prices superior to the first stage not through the grouping of the real estate prices predicted results. 2、Variables affect the real estate prices through the adaptive resonance theory clustering, the clustering results can affect the real estate as an alternative variable. 3、the best of the adaptive resonance theory clustering and the vigilance test parameter is set to 0.5, loading the second stage the Back-Propagation Network to estimate real estate prices. 4、The empirical results can be found, the back-propagation neural network input layer, remove abnormal samples were better, so the input variable, the sample should be removed outliers.
為了持續優化網站功能與使用者體驗,本網站將Cookies分析技術用於網站營運、分析和個人化服務之目的。
若您繼續瀏覽本網站,即表示您同意本網站使用Cookies。