Modelo de regressão linear múltipla: uma aplicação em dados de imóveis da cidade de Toledo - PR

Abstract

Statistics is a strong data management tool related to everyday situations and has contributed signicantly to the decision-making process in the most different areas as engineering, economics, medicine, among others. In the ramications of statistics, the study of Linear regression is one of the broadest techniques used that expresses the relationship between two or more variables, such that one variable can be predicted from others. The present work presents a multiple linear regression model for a dataset about price and characteristics of real estate located in the city of Toledo in Paraná State. The data were obtained through websites of several real estate agencies in the city. We consider as variables that influence the price of the house: built area, lot size, number of bathrooms, number of parking spaces and the neighborhood where the property is located (dummy variable). The modeling was performed using the software R and the stepwise method applied to select the best set of variables that describes the response variable, ie the price of the house. The validation of the optimal model was performed and it can explain 83% of the house value variability. Thus, the present work is a contribution to the real estate market, since the price of a house in the city of Toledo-PR can be estimated based on its characteristics.