基於分析個人偏好的房地產推薦系統

Real Estate recommendation system based on the analysis of personal preferences

吳建基(Chien-Chi Wu)

指導教授 ： 邱昭彰

元智大學/資訊學院/資訊管理學系/碩士(2012年)

摘要

隨著網路科技的發展與成熟，房仲產業在買屋、賣屋的相關資訊，也由早期的廣告看板、文宣和報紙轉變為現在的網路搜尋，顧客也能更快速且方便的獲取房屋資訊，但現行房仲產業的網路資訊提供，仍處於單純的資訊提供。當顧客不熟悉網路操作或購屋需求模糊時，將浪費許多時間在房屋的搜尋工作上。而Google、Bing、百度等搜尋引擎，則是偏重於全文檢索查詢，無法針對房屋特定需求提供有效的查詢機制。
本研究中，將發展一套房屋推薦系統，以全文檢索軟體Lucene為基礎，建立使用者的搜尋互動介面，並在使用者的房屋搜尋過程中，蒐集及記錄使用者的查詢行為，再以統計方法建立顧客的房屋特徵偏好模型，爾後便以偏好模型進行房屋推薦，在驗證效益的實驗中，實驗人員會分別以相同條件操作全文檢索查詢與推薦系統，實驗過程將分別記錄二者的點選紀錄，並以點選紀錄進行比較，經實驗證明，推薦系統的查詢效率較全文檢索提升43.36%。以此思維與方式，在未來，應可應用於其他傳統內容提供的網站，如求職、購物等網站，使更多網路使用者得到便利。

關鍵字

房地產 ； 個人偏好 ； 推薦系統 ； 搜尋引擎

並列摘要

With the network technology develops and matures, trading information of the real estate agencies follow the trend by the early advertising billboards, advertisements and newspapers transition to the current web search, so customers can be more rapid and convenient access to housing information. However, the websites’ search function of the current real estate agencies are simply providing information and it will waste a lot of time in searching while customers are not familiar with network operation or having fuzzy housing conditions. Large search engine, such as Google, Bing, Baidu…etc, emphasis in the full-text queries but cannot provide an effective search mechanism for specific housing needs.
This study develop a housing recommendation system based on the full-text search software Lucene with user interactive search interface, and able to collect and record a user's query behavior during housing search process. In this study, I establish the statistical model of the customer preference of housing characteristics, and provide housing recommendation according to the results. In order to verify the effectiveness, we use the same conditions in full-text search query and housing recommendation system, record and compare the times of click of these two systems. The empirical result shows that the effectiveness of housing recommendation system increase 43.36% compared with the original full-text search. Through this development, will be able to apply to other traditional websites such as job and shopping, and facilitate the process of searching users.

並列關鍵字

real estate ； personal preference ； recommendation systems ； search engine

參考文獻

2. 張怡文、江穎慧、張金鶚《分量回歸在大量估價模型之應用－非典型住宅估價之改進》都市與計畫 第 36 卷第 3 期 pp.281-304，2008。

Quality of Information for Wewb 2.0 Services. IEEE Internet Computing 14(6),

pp. 47-55, 2010.

basing recommendations on consumer product reviews," IEEE Intelligent

3. Billsus, D. and Pazzani, M., "User Modeling for Adaptive News Access,"Journal