Using attention networks to learn representations for house price prediction
Autor(a) principal: | |
---|---|
Data de Publicação: | 2019 |
Tipo de documento: | Dissertação |
Idioma: | eng |
Título da fonte: | Repositório Institucional da UFPE |
Texto Completo: | 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. |
id |
UFPE_fd346496a2870a122c113421ec5aa61b |
---|---|
oai_identifier_str |
oai:repositorio.ufpe.br:123456789/36780 |
network_acronym_str |
UFPE |
network_name_str |
Repositório Institucional da UFPE |
repository_id_str |
2221 |
spelling |
VIANA, Darniton Amorimhttp://lattes.cnpq.br/0517348203643422http://lattes.cnpq.br/7113249247656195BARBOSA, Luciano de Andrade2020-03-04T18:53:25Z2020-03-04T18:53:25Z2019-07-26VIANA, Darniton Amorim. Using attention networks to learn representations for house price prediction. 2019. Dissertação (Mestrado em Ciências das Computação) – Universidade Federal de Pernambuco, Recife, 2019.https://repositorio.ufpe.br/handle/123456789/36780Estimating 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.Estimar o valor de mercado de um imóvel é importante para muitos negócios, tais como imobiliárias e empresas de concessão de empréstimos imobiliários. O preço do imóvel não depende apenas de suas características estruturais (e.g. área e número de quartos) mas também de sua vizinhança. Este contexto pode ser explicitamente capturado, por exemplo, através da coleta de imagens de satélite ou pontos de interesse na vizinhança, ou implicitamente através da observação dos preços dos imóveis vizinhos. Como coletar explicitamente o contexto espacial é usualmente custoso, neste trabalho nós estimaremos o preço de uma casa baseado unicamente em suas características estruturais e as característica e preços de sua vizinhança. Para capturar implicitamente o contexto espacial de um imóvel, nós propomos um mecanismo de atenção hibrido que pondera a vizinhança baseada em sua similaridade em termos de características estruturais e localização geográfica do imóvel. Para as características estruturais, nós aplicamos uma atenção baseada na distância euclidiana e, para a localização geográfica, nós implementamos uma camada de atenção baseada em um kernel de função de base radial. Esses mecanismos de atenção são usados por um regressor em uma rede neural para aprender um vetor que representa um imóvel: o house embedding. Este vetor pode então ser usado como um conjunto de características para algum outro regressor realizar a predição do preço do imóvel. Nós realizamos uma extensiva avaliação experimental em 5 diferentes conjuntos de dados reais que mostram que: (1) Regressores usando house embedding obtém, na maioria dos casos, os melhores resultados em todos os 5 conjuntos de dados; (2) o house embedding aprendido melhora a performance dos regressores avaliados em quase todos os cenários comparado com os resultados usando os atributos originais; (3) modelos que utilizam regressores simples semelhante à regressores lineares, usando house embedding, alcançam comparáveis resultados em relação a algorítimos mais competitivos, tais como Random Forest e XGboost; (4) Nossa solução obtém os melhores resultado quando comparado ao uso de pontos de interesse; (5) Nossa abordagem supera tradicionais modelos preditivos espaciais; e (6) nossa solução obtém resultados melhores que abordagens anteriores que utilizam Deep Learning para predição do valor do imóvel que utilizam uma estratégia mais custosa para capturar o contexto espacial.engUniversidade Federal de PernambucoPrograma de Pos Graduacao em Ciencia da ComputacaoUFPEBrasilAttribution-NonCommercial-NoDerivs 3.0 Brazilhttp://creativecommons.org/licenses/by-nc-nd/3.0/br/info:eu-repo/semantics/openAccessBanco de dadosAutocorrelação espacialInterpolaçãoUsing attention networks to learn representations for house price predictioninfo:eu-repo/semantics/publishedVersioninfo:eu-repo/semantics/masterThesismestradoreponame:Repositório Institucional da UFPEinstname:Universidade Federal de Pernambuco (UFPE)instacron:UFPEORIGINALDISSERTAÇÃO Darniton Amorim Viana.pdfDISSERTAÇÃO Darniton Amorim Viana.pdfapplication/pdf14947227https://repositorio.ufpe.br/bitstream/123456789/36780/1/DISSERTA%c3%87%c3%83O%20Darniton%20Amorim%20Viana.pdfef0a0f0073ddc18d16e3e2c0c7f9b5cfMD51CC-LICENSElicense_rdflicense_rdfapplication/rdf+xml; charset=utf-8811https://repositorio.ufpe.br/bitstream/123456789/36780/2/license_rdfe39d27027a6cc9cb039ad269a5db8e34MD52LICENSElicense.txtlicense.txttext/plain; charset=utf-82310https://repositorio.ufpe.br/bitstream/123456789/36780/3/license.txtbd573a5ca8288eb7272482765f819534MD53TEXTDISSERTAÇÃO Darniton Amorim Viana.pdf.txtDISSERTAÇÃO Darniton Amorim Viana.pdf.txtExtracted texttext/plain167631https://repositorio.ufpe.br/bitstream/123456789/36780/4/DISSERTA%c3%87%c3%83O%20Darniton%20Amorim%20Viana.pdf.txtf643f28ca8cc5ce8c384995e90ac0d37MD54THUMBNAILDISSERTAÇÃO Darniton Amorim Viana.pdf.jpgDISSERTAÇÃO Darniton Amorim Viana.pdf.jpgGenerated Thumbnailimage/jpeg1248https://repositorio.ufpe.br/bitstream/123456789/36780/5/DISSERTA%c3%87%c3%83O%20Darniton%20Amorim%20Viana.pdf.jpg62b95cf8239b707bef2678d7a33290b5MD55123456789/367802020-03-05 02:17:07.451oai:repositorio.ufpe.br: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ório InstitucionalPUBhttps://repositorio.ufpe.br/oai/requestattena@ufpe.bropendoar:22212020-03-05T05:17:07Repositório Institucional da UFPE - Universidade Federal de Pernambuco (UFPE)false |
dc.title.pt_BR.fl_str_mv |
Using attention networks to learn representations for house price prediction |
title |
Using attention networks to learn representations for house price prediction |
spellingShingle |
Using attention networks to learn representations for house price prediction VIANA, Darniton Amorim Banco de dados Autocorrelação espacial Interpolação |
title_short |
Using attention networks to learn representations for house price prediction |
title_full |
Using attention networks to learn representations for house price prediction |
title_fullStr |
Using attention networks to learn representations for house price prediction |
title_full_unstemmed |
Using attention networks to learn representations for house price prediction |
title_sort |
Using attention networks to learn representations for house price prediction |
author |
VIANA, Darniton Amorim |
author_facet |
VIANA, Darniton Amorim |
author_role |
author |
dc.contributor.authorLattes.pt_BR.fl_str_mv |
http://lattes.cnpq.br/0517348203643422 |
dc.contributor.advisorLattes.pt_BR.fl_str_mv |
http://lattes.cnpq.br/7113249247656195 |
dc.contributor.author.fl_str_mv |
VIANA, Darniton Amorim |
dc.contributor.advisor1.fl_str_mv |
BARBOSA, Luciano de Andrade |
contributor_str_mv |
BARBOSA, Luciano de Andrade |
dc.subject.por.fl_str_mv |
Banco de dados Autocorrelação espacial Interpolação |
topic |
Banco de dados Autocorrelação espacial Interpolação |
description |
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. |
publishDate |
2019 |
dc.date.issued.fl_str_mv |
2019-07-26 |
dc.date.accessioned.fl_str_mv |
2020-03-04T18:53:25Z |
dc.date.available.fl_str_mv |
2020-03-04T18:53:25Z |
dc.type.status.fl_str_mv |
info:eu-repo/semantics/publishedVersion |
dc.type.driver.fl_str_mv |
info:eu-repo/semantics/masterThesis |
format |
masterThesis |
status_str |
publishedVersion |
dc.identifier.citation.fl_str_mv |
VIANA, Darniton Amorim. Using attention networks to learn representations for house price prediction. 2019. Dissertação (Mestrado em Ciências das Computação) – Universidade Federal de Pernambuco, Recife, 2019. |
dc.identifier.uri.fl_str_mv |
https://repositorio.ufpe.br/handle/123456789/36780 |
identifier_str_mv |
VIANA, Darniton Amorim. Using attention networks to learn representations for house price prediction. 2019. Dissertação (Mestrado em Ciências das Computação) – Universidade Federal de Pernambuco, Recife, 2019. |
url |
https://repositorio.ufpe.br/handle/123456789/36780 |
dc.language.iso.fl_str_mv |
eng |
language |
eng |
dc.rights.driver.fl_str_mv |
Attribution-NonCommercial-NoDerivs 3.0 Brazil http://creativecommons.org/licenses/by-nc-nd/3.0/br/ info:eu-repo/semantics/openAccess |
rights_invalid_str_mv |
Attribution-NonCommercial-NoDerivs 3.0 Brazil http://creativecommons.org/licenses/by-nc-nd/3.0/br/ |
eu_rights_str_mv |
openAccess |
dc.publisher.none.fl_str_mv |
Universidade Federal de Pernambuco |
dc.publisher.program.fl_str_mv |
Programa de Pos Graduacao em Ciencia da Computacao |
dc.publisher.initials.fl_str_mv |
UFPE |
dc.publisher.country.fl_str_mv |
Brasil |
publisher.none.fl_str_mv |
Universidade Federal de Pernambuco |
dc.source.none.fl_str_mv |
reponame:Repositório Institucional da UFPE instname:Universidade Federal de Pernambuco (UFPE) instacron:UFPE |
instname_str |
Universidade Federal de Pernambuco (UFPE) |
instacron_str |
UFPE |
institution |
UFPE |
reponame_str |
Repositório Institucional da UFPE |
collection |
Repositório Institucional da UFPE |
bitstream.url.fl_str_mv |
https://repositorio.ufpe.br/bitstream/123456789/36780/1/DISSERTA%c3%87%c3%83O%20Darniton%20Amorim%20Viana.pdf https://repositorio.ufpe.br/bitstream/123456789/36780/2/license_rdf https://repositorio.ufpe.br/bitstream/123456789/36780/3/license.txt https://repositorio.ufpe.br/bitstream/123456789/36780/4/DISSERTA%c3%87%c3%83O%20Darniton%20Amorim%20Viana.pdf.txt https://repositorio.ufpe.br/bitstream/123456789/36780/5/DISSERTA%c3%87%c3%83O%20Darniton%20Amorim%20Viana.pdf.jpg |
bitstream.checksum.fl_str_mv |
ef0a0f0073ddc18d16e3e2c0c7f9b5cf e39d27027a6cc9cb039ad269a5db8e34 bd573a5ca8288eb7272482765f819534 f643f28ca8cc5ce8c384995e90ac0d37 62b95cf8239b707bef2678d7a33290b5 |
bitstream.checksumAlgorithm.fl_str_mv |
MD5 MD5 MD5 MD5 MD5 |
repository.name.fl_str_mv |
Repositório Institucional da UFPE - Universidade Federal de Pernambuco (UFPE) |
repository.mail.fl_str_mv |
attena@ufpe.br |
_version_ |
1802310745339199488 |