巨量資料 : 公開資料與房仲網的房價分析

Abstract

多數人購屋資訊來自親友介紹，房仲網，和實價登錄網。不過這些資料分散在不同地方，缺少直接比較的資訊。本實驗用新竹縣實價登錄資料來建立分析房價模型。理解資料後，將不相關的房屋過濾掉，比如，篩選掉商業用途的辦公大樓。使用K-means分群，得到房仲網的平均價格比開放資料高的結論。使用「坪數」與「屋齡」，來看房仲網和開放資料的比值，並找實際物件來驗證。 在Ubuntu上，安裝Apache 網頁伺服器和 MySQL資料庫來架設網站。使用HTML和PHP語言來編寫網頁。資料庫字集設定為UTF8-Unicode，來處理內容為中文的房價資料。定期自動到政府資料開放平台和Yahoo奇摩房地產網，取得不動產買賣實價登錄資料和房仲網資料，由Python處理後，自動匯入資料庫，申請網路空間，提供房價分析服務 。本服務提供： 顯示同縣市同坪數同屋齡房仲網與開放資料的比較資訊， 總價元比值「平均值」和「標準差」。加入Google Analytics來收集使用者的瀏覽行為 。最後使用問卷取得使用者的意見回饋。本實驗對於購屋的消費者，提供房屋議價空間的資訊。

Information for buying a house is from friends, real-agent-web and register-real-price data for most people. However ,those data are in different places.There are no direct comparison. I build a model by using register-real-price data of Hsinchu County.First ,I observe data and delete irrelevant data.For example ,I delete office buildings for commercial purposes. Second,I use K-means clustering to get the conclusion.The average price of real-agent-web is higher than the average price of register-real-price .Third,I calculate ratios of real-agent-web 's price to register-real-price's price by the conditions of 「square feet 」and「age of building」. Fourth,I find some real instances to support the experiment.Fifth,I install Apache and MySQL in Ubuntu and write HTML and PHP. I use the UTF-8 character set to process Chinese words in the house-price data. I write a shell script.It can get data from data.gov.tw and tw.house.yahoo.com termly and automatically. I write Python code to process data.The program imports them to the database automatically.I apply for a web space in order to provide the service that analyzes house prices.The system compares the house-price information of real-agent-web and register-real-price in same counties,「square feet」and「age of building」. It also shows mean and standard deviation of price's ratios.I use 「Google Analytics 」to observe user's browsing behavior .I get users' feedback by questionnaires. In conclusion,the analysis of house prices is useful for consumers.