MASS EVALUATION WITH GENERALIZED ADDITIVE MODEL AND GEOSTATISTICS (Atena Editora)

Abstract

This work presents a methodology for mass evaluation using Generalized Additive Models and geostatistics for urban land data in the city of Fortaleza. The Generalized Additive Model confirmed some hypotheses regarding the behavior of real estate market prices, such as an increase in average unit value when the lot is within a gated community, with a water supply available and when located on a paved road. With this model, the non-linearity of unit prices with the tested area, utilization index, distance to main roads, commercialization density, income and IPTU base value of the current PGV was also revealed. However, the interaction of the model's terrain coordinates was not sufficient to eliminate the spatial autocorrelation of the residues, so a spherical variogram was adjusted over them, with subsequent interpolation by ordinary kriging, obtaining a residue surface at be added to the final prediction. From this procedure, there was an improvement in all performance measures used in this work, mainly regarding the COD, which increased from 34.40% to 13.06% after interpolation, which fully meets the IAAO recommendations for evaluation models. in large scale. It is concluded that the proposed methodology, using Generalized Additive Models and interpolation of residues with geostatistical techniques, is very promising in mass evaluation, but must be used very sparingly, as measurement errors in the collection of market data can bias the final prediction.