Recommending property with short days-on-market for estate agency

Estates with short days-on-market (DOM), referring to the properties whose days on the active market are short, attract realtors to gain commission quickly in transactions. With the rise of the Internet, massive information of on-sale houses can be obtained online. It is a challenge for estate agencies to mine estates with short DOM from such massive information. In this paper, we proposed an estate with short DOM appraisal framework to automatically recommend those estates using transaction data and profile information crawled from websites. Motivated by the estimation process of domain experts, we first seek similar estates with the location, structure and market information of estates. Then, the listing prices of similar estates were used to calculate a price interval. To evaluate the proposed framework, we used two real-world datasets, which consist of 220,000 on-sale estates and 643 estates with manually evaluated values in Chongqing real estate market. The results show that the proposed framework can estimate accurately about 78% estates. Two agencies scored the list of recommendation, in which 88% estates are trustworthy. Compared to the classical hedonic model, our method can successfully deal with problems such as the negative effects of special estates and the lack of labels. Moreover, both the spatial and temporal characteristics of an estate are integrated into our framework. The effect of different parameter settings is also discussed.