-
Tipo do ITEM
-
pt-BR
Artigo Científico
-
Título do Artigo
-
pt-BR
Estimation Of The Unitary Cost Of The Square Meter Popular Housing In The City Of Manaus Based On The Most Important Inputs, Using Artificial Neural Networks
-
Descrição
-
pt-BR
Abstract: Civil construction is one of the most expressive sectors in the economy, development and employability in the national territory. It is considered one of the branches that demonstrates the expansion and wealth of a country, where social housing is directly linked to socioeconomic development. According to the IBGE, in 2020, Manaus had 653,618 homes, of which 348,618 are classified as subnormal agglomerations, that is, stilt houses and unhealthy occupations and/or difficult to access. It can be said that one of the major obstacles to the construction of low-income housing is the lack of predictability of the behavior of costs during the execution of the work. This factor is even more pronounced in subdivisions and housing complexes, that is, in mass production, due to quantity. In order to mitigate these challenges, a tool was developed, based on the concepts of RNA -Artificial Neural Network, which compiles 2 civil construction price databases and predicts the cost of popular housing based on the value of the main inputs. This network seeks to estimate the cost per square meter of construction of popular housing in the city of Manaus. MATLAB® software was used, where data from the CUB and INCC databases were compiled. The inputs used were those contained in the so-called “basic batch”, recommended by the CUB. Quickly and practically without cost, the developed tool can predict the price of the square meter of popular housing, in the city of Manaus, from the stipulation of the inputs. RNA was able to present a very strong correlation in the sources of its sample space, thus demonstrating that the databases, despite presenting different data collection and treatment, in addition to being elaborated by different institutes, present compatibility in their databases, which is reflected in the veracity and reliability of the data collected and processed. After estimating several statistical indices, it is clearly noted that this is a tool that has proven to be efficient and safe for estimating future costs.
Background: The final cost of a work is one of the determining factors for carrying it out, especially in the more popular classes, where money is scarcer and any financial estimation errors can make the completion of the building unfeasible. In Brazilian territory, there are databases that predict the cost for the present, but none that estimate the cost for the future. This study develops an RNA to simulate the future value of a popular building in the city of Manaus.
Materials and Methods: In this study, two databases were used as a national recognition database: the CUB and the INCC. The reference values were extracted from the databases, in monthly cadence, during the period from July 2009 to May 2022, to obtain an arithmetic medium for each item, thus allowing RNA simulation to predict the value of the square meter of the house.
Results: Values for: MSE, NRMSE, MAPE, SER, MAE, RMSE, Medium Percent Error, and Pearson Correlation. All show adequate and correlated results. However, it can be stated that the most expressive result was the MSE, which was 91.14%, characterizing a well-adjusted RNA.
Conclusion: The correlation between the data in the two databases was 91.14%, enabling the simulation of an artificial neural network with data from different sources and good accuracy.
-
Abstract
-
en
Abstract: Civil construction is one of the most expressive sectors in the economy, development and employability in the national territory. It is considered one of the branches that demonstrates the expansion and wealth of a country, where social housing is directly linked to socioeconomic development. According to the IBGE, in 2020, Manaus had 653,618 homes, of which 348,618 are classified as subnormal agglomerations, that is, stilt houses and unhealthy occupations and/or difficult to access. It can be said that one of the major obstacles to the construction of low-income housing is the lack of predictability of the behavior of costs during the execution of the work. This factor is even more pronounced in subdivisions and housing complexes, that is, in mass production, due to quantity. In order to mitigate these challenges, a tool was developed, based on the concepts of RNA -Artificial Neural Network, which compiles 2 civil construction price databases and predicts the cost of popular housing based on the value of the main inputs. This network seeks to estimate the cost per square meter of construction of popular housing in the city of Manaus. MATLAB® software was used, where data from the CUB and INCC databases were compiled. The inputs used were those contained in the so-called “basic batch”, recommended by the CUB. Quickly and practically without cost, the developed tool can predict the price of the square meter of popular housing, in the city of Manaus, from the stipulation of the inputs. RNA was able to present a very strong correlation in the sources of its sample space, thus demonstrating that the databases, despite presenting different data collection and treatment, in addition to being elaborated by different institutes, present compatibility in their databases, which is reflected in the veracity and reliability of the data collected and processed. After estimating several statistical indices, it is clearly noted that this is a tool that has proven to be efficient and safe for estimating future costs.
Background: The final cost of a work is one of the determining factors for carrying it out, especially in the more popular classes, where money is scarcer and any financial estimation errors can make the completion of the building unfeasible. In Brazilian territory, there are databases that predict the cost for the present, but none that estimate the cost for the future. This study develops an RNA to simulate the future value of a popular building in the city of Manaus.
Materials and Methods: In this study, two databases were used as a national recognition database: the CUB and the INCC. The reference values were extracted from the databases, in monthly cadence, during the period from July 2009 to May 2022, to obtain an arithmetic medium for each item, thus allowing RNA simulation to predict the value of the square meter of the house.
Results: Values for: MSE, NRMSE, MAPE, SER, MAE, RMSE, Medium Percent Error, and Pearson Correlation. All show adequate and correlated results. However, it can be stated that the most expressive result was the MSE, which was 91.14%, characterizing a well-adjusted RNA.
Conclusion: The correlation between the data in the two databases was 91.14%, enabling the simulation of an artificial neural network with data from different sources and good accuracy.
-
Língua do arquivo
-
pt-BR
Inglês
-
Data da Publicação
-
pt-BR
01/06/2023
-
Palavra-chave
-
pt-BR
Neural Network
-
pt-BR
Popular housing
-
pt-BR
Inputs
-
pt-BR
Price
-
Autores
-
pt-BR
Arlindo Rubens de Oliveira Frota
-
pt-BR
Manoel Henrique Reis Nascimento
-
pt-BR
Antônio Estanislau Sanches
-
Editora
-
IOSR Journal of Business and Management (IOSR-JBM)
-
Local
-
Manaus / Brasil
-
Áreas de Conhecimento
-
pt-BR
Otimização de Processos Industriais
-
Turma
-
pt-BR
Turma 01
