客觀標準化不動產估價之可行性分析──市場比較法應用於大量估價

A Feasibility Analysis of the Standardization Valuation Model-Applying the Sales Comparison Approach to the Automated Valuation Model

本文運用估價師進行市場比較法行為邏輯，並結合特徵價格理論，建立大量樣本的客觀標準化估價模型。估價模型建構，乃透過特徵價格模型中的標準化β係數以計算明科斯基距離（Minkowski metric），作為選取可比較標的依據；調整階段以模型中的β係數作為依據；權重則是按照調整階段之比例作為依據。經實證發現，以隨機抽樣三十次進行重覆試驗觀察模型準確度，其平均絕對百分比誤差為15.48%，標準差為0.0066；落在正負10%與20%的命中率分別為41.82%與71.65%，而其標準差分別為0.0268與0.0229。整體而言，模型準確度達一定水準，且具相當穩定程度，顯示透過標準化市場比較法進行大量估價是可行的。此外，運用此模型分析，估價師不需要主觀預測，可改善過去估價結果不一致情形，並達到大量估價目的。

The purpose of this paper is to examine an alternative to the sales comparison approach using the automated valuation model. To qualify the comparables, we adopt the Minkowski metric and β coefficient of the hedonic model to select comparable sales. The empirical results show that the MAPE of the AVM model is 15.48% based on thirty repeated experiments using random sampling. The hit rates within 10% and 20% are 41.82% and 71.65%, respectively, and the corresponding standard deviations are 0.0268 and 0.0229. The findings indicate that the application of the AVM model is feasible and could overcome the drawbacks of the appraiser's subjective bias.