有鑑於房地產景氣相關研究中市場需求面分析研究漸受到重視,其中家戶的購屋決策行為更是一個重要的研究課題。國內、外多篇研究文獻對於購屋決策行為,細分為多個可能具有影響力的購屋決策因素,大部分文獻採用迴歸模型、因素分析等統計方法對歷史資料進行分析或預測。然而,由於網際網路和電腦相關科技技術的蓬勃發展,企業的資料蒐集量常常以百萬、甚至千萬計,但是資料庫內有用且潛藏的知識卻難以利用統計方法發現。因此本研究發展一個二階段實驗方法(two-phase approach)對購屋決策因素深入探討。第一階段實驗使用資料探勘技術中的k-means演算法對已購屋者做市場區隔,再用關聯法則分析尋找潛藏的有趣樣式。第二階段實驗使用分析層級程序法進行購屋決策制定。本研究主要目的即為探討台北市部分行政區家戶購屋決策因素之關聯性異同與重要程度之排序,希望藉此幫助房地產業者對於具有關聯性的購屋決策因素有更深入的認識,並做為營建房地產市場策略擬定的參考,如此一來公司行銷策略之制定與有限資源的分配都能夠更臻合理,同時也能增加公司的競爭優勢。
In the real estate cycle study, the house buying decision making has been enormously investigated in recent years. Most literatures used regression models, factor analysis etc. to deal with historical real estate transaction data. However, as the Internet and computer technology booming affected significantly the framework of data collection, the transaction data in modern trading environment always contain millions (or more) records. Discovering potential interesting patterns in databases by statistical techniques such as regression model or factor analysis is indeed difficult. Obviously, developing an analytical technique for such enormous transaction data in databases is certainly taken into account. This study develops a two-phase approach to help make house buying decision. In phase one, we use the k-means algorithm, a well-known clustering method in data mining, to do customer-segmentation, then search for potential interesting patterns and house buying factors by associating rule mining. In phase two, the Analytical Hierarchy Process is used to explore house buying decisions with the factors discovered in phase one. This study intends to provide related house buying decision factors and its corresponding ranking for further applications. The real estate agents may check the conclusion of this study as the reference of decision making in real estate market.