Modelagem preditiva do preço de aluguel de apartamentos por bairros na cidade de São Paulo
Data
2021-08-12
Autores
Ratcow Junior, Gerson 
Orientadores
Lucas, Edimilson Costa
Tipo
Trabalho de conclusão de curso
Título da Revista
ISSN da Revista
Título de Volume
Resumo
Este estudo tem como objetivo a aplicação de Machine Learning para realizar
uma modelagem preditiva dos preços de aluguel de apartamentos mapeados na
cidade de São Paulo, a fim de que seja identificado as variáveis que mais impactam o
preço de aluguel. Para realizar a análise foi utilizada uma base de dados de anúncios
de imóveis para alugar na cidade de São Paulo em abril de 2019. O método de
modelagem foi testado em três algoritmos, são eles: Random Forest Regression,
XGBoost e Suport Vector Regression. Apesar dos resultados positivos obtidos em
todos os algoritmos, foi constatado que o algoritmo XGBoost apresentou melhor
acurácia na modelagem do presente estudo. Os resultados da modelagem indicam
que as variáveis mais relevantes na composição do preço de aluguel dos
apartamentos da cidade de São Paulo são: tamanho do apartamento, localização e
área de lazer com piscina.
This study aims to use the Machine Learning application to perform a predictive modeling of average and real rental prices of mapped apartments in the city of São Paulo, in order to identify the variables that most impact the rental price. To perform the analysis, a database of property advertisements that were for rent in the city of São Paulo in April 2019 was used. The modeling method was tested in four algorithms, they are: Random Forest Regression, XGBoost and Support Vector Regression. Despite the positive results obtained in all algorithms, it was found that the XGBoost algorithm showed better accuracy in the modeling of the present study. The model results indicate that the most relevant variables in the composition of the rent price of apartments in the city of São Paulo are apartment size, location and leisure area with swimming pool.
This study aims to use the Machine Learning application to perform a predictive modeling of average and real rental prices of mapped apartments in the city of São Paulo, in order to identify the variables that most impact the rental price. To perform the analysis, a database of property advertisements that were for rent in the city of São Paulo in April 2019 was used. The modeling method was tested in four algorithms, they are: Random Forest Regression, XGBoost and Support Vector Regression. Despite the positive results obtained in all algorithms, it was found that the XGBoost algorithm showed better accuracy in the modeling of the present study. The model results indicate that the most relevant variables in the composition of the rent price of apartments in the city of São Paulo are apartment size, location and leisure area with swimming pool.
Descrição
Citação
Ratcow, Gerson Junior