Applying Model Trees on Actual Registration Data to Filter Critical Factors for the Variation of Real Estates

Abstract

The government’s real-value registry policy for residential justice provides more information for people to assess the values of real estates This policy could reduce the information asymmetry between buyers and sellers Many applications have been developed for this kind of open data to filter useful information for the market of real estate This study collects the real-value registry data from years 2016 to 2018 located in Tainan city to determine the critical attributes for the up and down percentage of real estates The objects of real estimate are divided into three categories: land townhouse and apartment Since the address of an object is not fully revealed the objects with the same area located in the same section are used to estimate the up and down percentage of its value Then the attributes with strong collinearity are removed The remaining attributes are employed to grow a model tree year by year for class prediction and result interpretation The experimental results suggest that the period of hold has the most significant positive impact on the value of an apartment The size of land area has a positive impact on the value of a townhouse while the impact of the size of its floor area is negative The age of a townhouse is also an important attribute for its value No explicit conclusions can be made for lands These results can be helpful for people to estimate the prices of real estates and to recognize the usefulness of open data

Date of Award	2020
Original language	English
Supervisor	Tzu-Tsung Wong (Supervisor)