基於政府開放資料的房價預測系統

Abstract

買房子對一般人而言是一筆很大的開銷，故在買房前作謹慎評估是有其必要的。若能在買房前先對房價進行預測，以得知未來的房價為何，這對買房者來說是相當重要的參考資訊，如此一來即可根據預測房價，來決定是否要買該棟房子或當作議價的參考。 隨著公開資料的普及，買賣雙方資訊不對等的情況漸漸消失，各種透過分析房地產交易資料，再視覺化分析結果的服務紛紛出現。但這類型的服務大多是分析現有資料，再以圖表呈現歸納結果，提供房價預測的服務並不常見。目前所見的房價預測服務，預測方法通常較為簡單，僅僅使用統計或是簡單的機器學習模型去預測，只能在網站上得知預測的房價。 本研究提出了一個房價預測系統，針對給提供房價預測服務者使用，服務提供者可經由我們的系統，預測出更精準的房價，以提供更好的服務給一般買房者。使用者能自己客製化預測方式，並在GUI上透過文字以及圖表得知預測結果。我們的預測方式不同於以往，不僅僅是單純地透過預測模型作預測，而是先透過分群，將同一類型的房子歸在同一群後，於每一群分別作預測再統合預測結果。此外，本系統能對處理過的資料保存其分析結果，這意味著每當有新進的交易資料時，只需花時間在處理新資料上，如此可大幅降低預測房價的時間。而透過預測已知月份的房價，結果也顯示該系統確實能準確地預測房價。

It is a great expense for most people to buy the house. Thus, it is necessary to assess the information related to house carefully, especially the cost. If we can predict the house prices before buying it, it would be the important information for home buyers. They can decide whether to buy the house and bargain house price according to the prediction result. With the popularity of open data, the situation of information asymmetry between buyers and sellers disappears gradually. The service via data analysis and data visualization emerges in recent years. Most services analyze the data and present the result by plot rather than predict house price. The prediction methods are simpler in existing house prices prediction service. They merely use statistics or simple machine learning models to predict house prices. This thesis proposed the house prices prediction system for service providers. They can provide better service to home buyers by predicting more precise result through our system. User can select the prediction method and get prediction result from GUI. Unlike past analytical method we cluster data, make the prediction for each cluster and integrate all results. Besides, the system can preserve the analytical result. It is an useful feature when new transaction data are generated. The time of house prices prediction would be reduced dramatically because we only need to analyze new data rather than historical data. By predicting the past known house prices, we can know that the system can indeed predict house prices precisely.