全台區域房價預測網路整合服務

Abstract

在台灣 , 買房子對於現在的人來說並不是一件容易的事 , 常常需考慮很多因素 , 如房子本身的外表因素以及地理因素等 , 這些因素可能會直接或間接影響房子本身的價值。買房者在買房子之前 , 可能會透過網路查詢相關待售屋資料 , 目前網路上有許多房屋待售網站如永慶房屋網、東森房屋網、信義房屋網等 , 而買房者最常關注的地方是他們所想要的房屋類型的價格趨勢。 有鑒於目前的待售屋網站只提供目前的待售屋價格和詳細資訊 , 並無提供該類型的房屋的價格趨勢 , 網路上缺乏此種整合式的服務。並隨著台灣行政院內政部公開的實價登入資料漸漸普及及透明 , 許多待售屋網站僅利用實價登入資料去統計目前該類型房屋的平均價格 , 鮮少對該類型做合適的房價預測。 因此本系統提出一個整合房價預測和現有待售屋資訊的網路服務 , 買房者透過此服務可以選擇所喜好的房屋類型 , 利用本系統的一系列分析方法和預測模型來對該類型的房價做合理預測。在實驗結果上我們利用命中率來驗證我們給出的預測區間是否合理 , 六都中命中率在75%以上佔大半數 , 表示我們的分析預測方法有一定的效果。最後將預測結果以網頁呈現 , 並把現有符合該類型的房屋顯示在網頁上供買房者觀看 , 而我們也對30位受訪使用者做SUS滿意度調查表 , 結果上介於”GOOD”和”EXCELLENT”之間 , 顯示本系統具有一定的可用性。

Buying a house for most people is not an easy thing in Taiwan. Therefore, they have to consider many factors. For houses’ appearance factors and for geographical factors, these factors will effect directly or indirectly effect the value of house itself. Before buying house, house buyers maybe use Internet to inquire the information of house for sale. And there are many websites of house for sale on the internet such as Yungching.com, Etwarm.com, Xinyi.com, etc. For house buyers, they always concern about the trend of house price of the house they want. Because of the websites of house for sale only providing the current price of houses and details of it, it doesn’t provide the trend of price for specific house type. Hence, it lacks of the integrated service on internet. Along with the Actual Price Registration Data that Dept. of Land Administration announces becomes more popular and transparent than before, many web-sites of house for sale only take it to do the statistics for the average price of specific house type in current, not for the prediction of the house price. Consequntly, our system propose an integrated web service of house price prediction and current information of houses for sale. They can choose the house type they want by our web service, and utilize our system to do a series of analytical methods and prediction model to do the prediction of the house price. In the experiment result, we use hitrate to validate the legitimacy of our prediction interval we produce. For the six municipalities, there is about half of the hitrate above 75%. It implys that our analysis of prediction method has some effect. At last we will present the webpage of prediction result and show the information of houses for sale that in accordance with the specific house type that house buyers choose. We also choose 30 interviewers to do the System Usability Scale of our system, and the result is between “GOOD” and “EXCELLENT”, it indicates that this system is usable.