Boosting house price predictions using geo-spatial network embedding

Real estate contributes significantly to all major economies around the world. In particular, house prices have a direct impact on stakeholders, ranging from buyers financing companies. Thus, plethora of techniques been developed for real price prediction. Most existing rely different features build variety prediction models predict prices. Perceiving effect spatial dependence prices, some later works focused introducing regression improving performance. However, they fail take into account geo-spatial context neighborhood amenities such as how close is train station, or highly-ranked school, shopping center. Such contextual information may play vital role in users’ interests and thereby has influence its price. this paper, we propose leverage concept graph neural networks capture house. present novel method, network embedding (GSNE), that learns embeddings houses various types points interest (POIs) form multipartite networks, where POIs are represented attributed nodes relationships between them edges. Extensive experiments with large number show produced by our proposed GSNE technique consistently improve performance task regardless downstream model. Relevant source code available at: https://github.com/sarathismg/gsne .