Modelagem dos preços de imóveis residenciais no município de Aracaju via mistura de modelos de regressão

Abstract

The Real Estate Market Observatory of Aracaju (OMI-AJU), linked to the Municipal Finance Department of Aracaju (SEMFAZ-AJU), is dedicated to collecting and analyzing data on the real estate market in Aracaju. One of the objectives of OMI-AJU is to establish regression models that allow explaining and predicting the price of a property, given some of its characteristics, such as, for example, the neighborhood where it is located, the level of infrastructure of the place, its age, the number of parking spaces it has, etc. Using this data, an analysis was performed and a linear regression mixture model was developed, using the EM algorithm and the Bayesian Criterion (BIC) to select the number of groups (models) based on the characteristics of the properties. After the analyses, it was found that the highest value of a property in Aracaju, based on the information available, was R$ 39,000,000. This value was divided into three types of record: oer, transaction and ITBI, with ITBI containing the most observations, with approximately 36.41%. Regarding the regression model mixing technique, using three repetitions and a maximum number of 10 groups, through the BIC criterion, 5 groups (models) were dened. Group 5 was the one with the highest probability of an observation belonging to it, with just over 34%, while group 3 had the lowest probability, approximately 2.77%. Through the scatter plot of the groups, using latitude and longitude, it was observed that the distribution of the groups is well spread, indicating a high variability between the groups.