Using Text Mining To Analyze Real Estate Classifieds
This work addresses the lack of analysis in real estate classifieds for brokers and analysts, though it is incremental as it applies existing text mining techniques to a new domain.
The paper tackles the problem of predicting real estate prices from classified ads by proposing a two-stage regression model that exploits textual data, achieving significantly lower root mean squared error across multiple datasets compared to other regression models.
Many brokers have adapted their operation to exploit the potential of the web. Despite the importance of the real estate classifieds, there has been little work in analyzing such data. In this paper we propose a two-stage regression model that exploits the textual data in real estate classifieds. We show how our model can be used to predict the price of a real estate classified. We also show how our model can be used to highlight keywords that affect the price positively or negatively. To assess our contributions, we analyze four real world data sets, which we gathered from three different property websites. The analysis shows that our model (which exploits textual features) achieves significantly lower root mean squared error across the different data sets and against variety of regression models.