| Summary: | 碩士 === 國立成功大學 === 工業與資訊管理學系碩士在職專班 === 102 === Real estate appraisal is a complex task owing to the widely differing geographic factors of residential properties. If a bank does not hire a professional appraiser, it might become a subject to higher risk. By means of data mining, this research utilizes the real estate appraisal data of a bank to induce a model for predicting housing prices. Another predictive model learned from the actual selling prices collected from the Department of Land Administration, which is a part of the Ministry of the Interior in Taiwan, is also generated for comparison. The experimental results show that model tree can have a less mean absolute error than linear regression. Moreover, both actual selling prices and the bank appraisal depend on location, road conditions, and the total area of a house. A greater emphasis is placed on the building type for determining the actual price, while bank appraisals place a greater emphasis on the age of a building.
|
|---|
