Detalhes bibliográficos
Ano de defesa: |
2019 |
Autor(a) principal: |
VIANA, Darniton Amorim |
Orientador(a): |
BARBOSA, Luciano de Andrade |
Banca de defesa: |
Não Informado pela instituição |
Tipo de documento: |
Dissertação
|
Tipo de acesso: |
Acesso aberto |
Idioma: |
eng |
Instituição de defesa: |
Universidade Federal de Pernambuco
|
Programa de Pós-Graduação: |
Programa de Pos Graduacao em Ciencia da Computacao
|
Departamento: |
Não Informado pela instituição
|
País: |
Brasil
|
Palavras-chave em Português: |
|
Link de acesso: |
https://repositorio.ufpe.br/handle/123456789/36780
|
Resumo: |
Estimating the market price of a house is important for many businesses such as real estate and mortgage lending companies. The price of a house depends not only on its structural features (e.g. area and number of bedrooms) but also the spatial context where it is located. This context can be explicitly captured, for instance, by collecting satellite images or points of interest in the neighborhood, or implicitly by looking at the price of the nearby houses. Since collecting explicit spatial context is usually costly, in this work we estimate the price of a house based solely on its structural features and the characteristics and price of its neighbors. To capture the implicit spatial context of a house, we propose a hybrid attention mechanism that weights neighbors based on their similarity in terms of their structural features and geographic location to the house. For the structural features, we apply an euclidean-based attention and, for the geographic location, we implemented an attention layer based on a radial basis function kernel. Those attention mechanisms are used by a neural network regressor to learn a vector representation of the house defined as the house embedding. This vector can then be used as a feature set by any regressor to perform house price prediction. We have performed an extensive experimental evaluation on 5 different real-world datasets that shows that: (1) regressors using house embedding obtained, in most cases, the best results on all 5 datasets; (2) the learned house embedding improves the performance of the evaluated regressors in almost all scenarios comparing to their results using raw features; (3) simple regressor models such as Linear Regression using house embedding achieved comparable results to more competitive algorithms, such as Random Forest and XGboost; (4) Our proposed solution obtains better results about the use of points of interest; (5) Our approach outperformed traditional spatial predictive models; and (6) our proposed solution outperformed previous Deep Learning approaches for house price prediction that use more costly strategies to capture the spatial context. |