Capturing locational effects: application of the K-means clustering algorithm

This study proposes a hedonic pricing model to efficiently capture the values of locations without assuming a specific functional form or the factors affecting it. The K-means clustering algorithm serves as a subdivider for allocating indicators into samples according to their locational similarity. The advantage of this approach is that it allows the value of a location to be measured using only its latitude and longitude. We examine the predictive accuracy of the model in an out-of-sample context based on apartment transaction data for Seoul, the capital of Korea. Our results show that the predictive power of the proposed model is significantly higher than that of conventional models.