基於深度學習結合經濟指標之房價預測

2020年以來，COVID-19疫情於各地肆虐，不僅對全球經濟造成嚴重衝擊，在疫情急促上升的狀況下，房價也預期將受到影響。然而，疫情至今已長達兩年，房價目前並沒有如2003年遇上SARS時的跌落，反之由於前期防疫成效顯著、低利環境，以及資金回流等原因，使得房價不降反升，故本研究將透過實價登錄房價資料及總體經濟指標，使用不同深度學習方法來預測房屋總價，目的是為了探討疫情前後房地產的變化及對總體經濟產生的交互關係，最後可供房屋買家、賣家，以及租戶和房東參考，能以此為依據評估房屋價格是否合理。本研究實驗結果顯示，加入經濟指標變數對於房價預測正確率的改善有限。原因可能為，在疫情時期，由於低利率、儲蓄金額增加、以及通貨膨脹等原因，降低了買房的門檻，造成房價變得更難以預測，未來研究可再嘗試其他方法或變數以因應此情況。

關鍵字

房價預測；房價分析；深度學習；總體經濟；實價登錄

並列摘要

Since 2020, the COVID-19 has been raging in various places. It has caused a serious impact on the global economy. Besides, under the situation of a rapid rise in the pandemic, housing prices are also expected to be affected. However, it has been two years since the pandemic started, yet house prices have not fallen as they did in the SARS outbreak in 2003. On the contrary, due to the remarkable results of prevention in the early stages, the low interest environment, and the return of funds, housing prices have not fallen but risen. In this study, different deep learning methods and general economic indicators are used to predict housing prices. By exploring the changes in real estate at different stages of the pandemic, and the interaction with the overall economy. The proposed method can be used as a basis to assess whether a transaction is reasonable. The experimental results show that the improvement on the accuracy of predicting housing prices by adding economic indicator variables is limited.Low interest rates, increased savings, and inflation during the pandemic have lowered barriers of buying a house which might be the reasons that came with the results. Therefore. the housing prices have become more unpredictable. Future researches can explore more methods or variables to respond to this situation.