For the valuation of residence real estates, valuators use the Market Data Approach to compare the current case with other precedents of similar conditions to adjust the predicted price. Therefore, the accuracy of valuation depends on the valuator’s experience and ability to collect precedents. However, prices are hard to collect in the practical market. Inaccuracies in quotations occur frequently due to insufficient data or the variance in the experience of each valuator.
This study uses Case-Based Reasoning (CBR), which collects Real Estate factors as problem attributes through interviews with experts to describe the characteristics of each precedent and construct the case base. Eigenvector questionnaires are used to convert the experts’ evaluation of factors into relative weights. At last, the system finds a number of similar cases and filters them with the model proposed by this study to determine the effectiveness of each case in order to improve the accuracy of valuation.
The unit and total prices are computed respectively by selecting 3 verification cases for each residence type and filtering out the 7 most similar precedents. The result of this study shows that the accuracy is higher for general real estates than those on the main line or with penthouse-surcharges. Such result verifies that such approach is practical and accurate to general residence real estates in the region given in this study, while those on the main line or with penthouse-surcharges should be viewed as special housing submarkets with the process of problem attribute selection, weight conversion and similarity definition redone to form their individual models.
