基於廣義加成模型之房屋價格預測方法

Housing Price Forecasting by Generalized Additive Models

黃士峰(Shih-Feng Huang) ； 廖育苹(Yu-Ping Liao)

《中國統計學報》 60卷2期 (2022/06) Pp. 95-124

摘要

本研究針對不動產實價登錄資料，提出一結合廣義加成模型（generalized additive model，GAM）的系統性作法預測2014到2019年間高雄市的房屋價格，其中資料欄位包含數值型與類別型的解釋變數，例如：房屋位置、面積、型態…等。我們提出以GAM捕捉解釋變數對房價的非線性影響趨勢，並透過使用AIC（或BIC）決定哪些解釋變數需要透過非線性轉換方可提升模型的配適及預測表現。在實證研究方面則透過移動視窗法，以每3、6、12、或24個月的資料進行模型配適，再將其應用於預測下1個月的房屋價格，以評估所提出模型的表現。特別地，我們採用機器學習領域中的K最近鄰（K nearest neighbor，KNN）方法所估計的房屋價格作為比較基準。數值結果顯示所提出的模型對高雄市房屋價格的配適與預測表現皆優於KNN法。

關鍵字

廣義加成模型 ； 實價登錄 ； 薄板樣條

並列摘要

This study proposes a systematic approach based on the generalized additive model (GAM) and the real price registration website data to predict the housing prices in Kaohsiung from 2014 to 2019. The data include numeric and categorical covariates such as house location, area, type, etc. We employ the GAM to capture the nonlinear effects of covariates on housing prices, where the AIC (or BIC) is used to determine which covariates need to be transformed nonlinearly to improve the fitting and prediction performances. In our empirical study, a rolling window approach with a window size of 3, 6, 12, or 24 months and a one-month adaption frequency is employed to investigate the performance of the proposed method. In particular, we adopt the housing prices estimated by a machine learning method, the K nearest neighbor (KNN), as a comparison benchmark. The numerical results reveal that the proposed GAM has better fitting and predicting performances for Kaohsiung housing prices than the KNN method.

並列關鍵字

generalized additive models ； real price registration ； thin plate spline

參考文獻

Altman, N. S. (1992). An introduction to kernel and nearest-neighbor nonparametric regression. American Statistician, 46(3), pages 175-185.

Cunningham, P., and Delany, S. J. (2021). k-Nearest neighbour classifiers-A Tutorial. ACM Computing Surveys (CSUR), 54(6), pages 1-25

Dbrowski, J., and Adamczyk, T. (2010). Application of GAM additive non-linear models to estimate real estate market value. Geomatics and Environmental Engineering, 4(2), pages 55-62.

De Souza, J. B., Reisen, V. A., Franco, G. C., Ispány, M., Bondon, P., and Santos, J. M. (2018). Generalized additive models with principal component analysis: an application to time series of respiratory disease and air pollution data. Journal of the Royal Statistical Society: Series C (Applied Statistics), 67(2), pages 453-480.

Dominici, F., McDermott, A., Zeger, S. L. and Samet, J. M. (2002). On the use of generalized additive models in time-series studies of air pollution and health. American Journal of Epidemiology, 156(3), pages 193-203.