買賣實例比較法實例權重之分析與模式建立

Analysis and Model for Weights Used in the Sales Comparison Approach

買賣實例（市場）比較法為不動產估價的三種方法之一，一般需蒐集數個比較標的，經調 整後之試算價格決定比較價格。國外以數學計算式計算實例權重雖已相當普遍，但目前尚無應 用特徵效用模式，解讀實例權重形成與比較標的內部條件之關聯。本文設定特徵權重模型，發 現比較法買賣實例權重受價格型態、交易日期接近性、是否屬近鄰地區、實例差異百分率絕對 值加總、實例比較項目修正數、其他兩個實例相對值等自變數影響顯著。以 96 年各直轄市及 縣市政府蒐集之 3,114 筆權重實例中，隨機抽取 90%推估樣本外 10%之實例權重結果，於重 複測試十次結果，MAPE 甚小，且 Hit Rate 於 20%內甚高，符合本文設定之預測條件。本文直 接預測實例權重探討權重問題核心，且對於實例權重模型兼顧所分析實例本身以外之其他實例 相對條件，實證結果能使預測準確度提高之發現，可供未來不動產估價實務作業時，決定加權 平均賦予實例權重之參考，俾提高估價科學化及客觀程度。

The sales comparison approach is one of three real estate approaches used to appraise property value. Since one must collect several comparables, the weighted average of comparables for sales comparison value is commonly derived using a mathematical formula. However, weights will not necessarily be supported by the market when derived using only a mathematical formula. After collecting 3,114 sales comparables with weights for each city and county in Taiwan, this study examines the correlation between weights and internal attributes of 3,114 sales comparables using the hedonic price model. Price, the proximity to transaction date, inside a neighborhood area, total gross adjustment (%), number of adjustments, and attributes of other comparables in one appraisal significantly impact the weights of comparables. The expected mean absolute percentage error (MAPE) and Hit rate criteria are met after forecasting 10% validation samples modeled by 90% of randomly surveyed samples.