不動產自動估價與估價師個別估價之比較──以比較法之案例選取、權重調整與估值三階段差異分析

A Comparison of Automated Real Estate Valuation Models and Appraisers' Appraisals-Comparables Selection, Weighting Adjustment and Valuation for the Sales Comparison Approach

不動產估價研究可分為個別估價與大量估價兩類，過去有關個別估價研究多為估價行為或估價方法改進，而大量估價研究則多是運用數量方法建立模型，由於兩者建構方式不同，欲同時進行兩者實證比較並不容易。本文藉由相同勘估標的，取得估價師個別估價資料，與資料庫建立的自動估價系統，進行案例選取、權重調整及最後估值三階段的比較分析。經本文差異比較發現，自動估價系統選取比較案例差異小於個別估價，且自動估價系統符合比較案例相似度高時，給予其比較價格相對較高權重原則，從行為估價觀點，結果顯示自動估價系統具有較客觀且符合估價理論程序優點。然，受限於自動估價系統樣本資料分配限制，在資料樣本較少地區以及非典型住宅類型（面積過大或過小），比較結果為個別估價估值與自動估價系統估值差異大，顯示自動估價系統有其適用限制，未來若增加資料庫樣本或次市場模型，兩者估值差異將可獲得改善。

In real estate appraisal research, appraisers and mass appraisals are the two methods most often used in the sales comparison approach. In the past, most appraisers' studies focused on the appraisers' behavior or on improving appraising methods. Since these two methods are constructed in different ways, it is not easy to find enough objects for the analysis and a comparison between appraisers and mass appraisal. The main purpose of this paper is to select certain target cases and then compare the data provided by appraisers as well as the results of mass appraisals. The analysis is conducted in three stages: selecting comparables, adjusting weights and estimating values. The differences in the analysis indicate that the differences in the Automated Valuation Model (AVM) of the Minkowski metric of the comparables is smaller than the appraisals provided by individual appraisers. The findings indicate that the application of the AVM model is more objective and fits the appraisal procedure. However, with the limits of the AVM, the sparseness of the data and non-typical housing, the results of the AVM in some areas differ from those made by appraisers, which indicates that the AVM has limitations. This situation is expected to improve as a result of accumulating more data and adjusting the models. If the accuracy of the AVM is assured, the AVM will be applied more widely.