類神經網路應用於房地產估價之研究Neural Networks Technique for Residential Property Appraisal in Taipei

由於房地產市場屬一不完全市場,消費者在購屋的同時常因資訊的不健全而遭受不必要之損失,因此如何更精確的獲得房地應估價資訊賞局健全台灣房地應市場之要務。本研究探討如何有效地感用類神經網路於房地產估價上類神經網路與房地產估價結合之研究在國內仍未有相關文獻發表,然近年來在國外已有不少相關研究相繼發表。為了解探討如何有效地應用頻神經網路於房地產估價上,本研究設計了四個實驗,分別對不同的資料型態進行測試,並比較倒傳遞類神經網路(BP)、理解倒傳遞類神經網路(RNBP)輿廻歸方式特徵價格法三種模式於估價精確度上之差異,以烏未來改進房地產估價輔助系統之參考。研究結果顯示類神經網路透過適當的樣本學習,的確能獲得較好的估價結果,其中又以RNBP較BP為佳相較於傳統迴歸方式特微價格法,類神經網路未來應可拓展於房地產估價上。
(關鍵詞:房地產估價、特徵價格法、倒遞類經網路、解倒傳遞類神經網路)

This rescarch explores the concerns associated with applying Artificial Neural Networks (ANN)technology to real cstate appraisal, as well as compares the performances of two ANN models with a traditional multiple regression model in estimating the sale price of residential properties in the Taipei metro area. The study is based on 2787 sales of homes in Taipei. Significant concern is the requirement of sufficiently good training data and a better tearning algorithm. Every appraiser who plans on using ANN should do with such caution. Our experimental results do support previous findings that ANN is a superior tool for appraisal analysis, and show that the RNBP neural networks outperforms the back propagation neural networks.
Key words: Artificial Neural Networks (ANN), Housing, Appraisal, Taipei