Acceso usuarios
Pei Wang, Mingchin Chen
While many studies have applied data mining techniques to judge housing prices, few have decoded the important attributes or prioritized them simultaneously. This paper aims to utilize five data mining techniques to discover the important attributes for three major types of real estate in Taipei city. The datasets, involving a total of 22,480 transactions, were publicly available from the Taiwan Actual Price Registration from July 2013 to August 2015. The five models are decision trees, random forests, model trees, artificial neural networks and multiple regression. The criteria used to measure the forecasting accuracy are MAPE, R², RMSE, MAE and COR. The model with the best performance for all houses is the Model Tree with a MAPE value of 27.59. As for apartments, the best is Random Forests. Artificial Neural Networks perform best for suites and buildings with elevators. Different housing types need different models. Furthermore, the attributes importance helps us to conclude the really critical attributes, which include the floor area, administrative districts, parking area and land area, and their rankings. This variable ranking and selection procedure proposed by this research can also be adopted to improve the prediction efficiency for most big data applications other than the housing transactions.
