Using Data Mining Techniques to Model Housing Rental Price near Universities in Seoul

The main motivation of this research is to help university students who are seeking for their residential places, by providing objective information based on data. To this end, we gathered data for a large selection of rental units from Zigbang which is one of the most popular real estate mobile applications in South Korea. Additional information such as distance-to-school-gate which is unavailable from the mobile app was included in our analysis for the purpose of building more accurate models. We employed ridge regression, neural networks, support vector regression, and random forests to model housing rental price based on about 120 thousands observations. The trained models showed the prediction accuracy at around 96%. We also attempted to find out which factors are the most influential in pricing rental fees by analyzing interpretable models.