See discussions, stats, and author profiles for this publication at: https://www.researchgate.

net/publication/367227464

Noise estimation using an artificial neural network in the urban area of Jaen,
Cajamarca

Article in Bulletin of Electrical Engineering and Informatics · January 2023

DOI: 10.11591/eei.v12i3.4633

CITATION READS

1 201

4 authors, including:

Lenin Quiñones Huatangari

National University Toribio Rodríguez de Mendoza
58 PUBLICATIONS 178 CITATIONS

SEE PROFILE

All content following this page was uploaded by Lenin Quiñones Huatangari on 18 January 2023.

The user has requested enhancement of the downloaded file.

Bulletin of Electrical Engineering and Informatics
Vol. 12, No. 3, June 2023, pp. 1427~1434
ISSN: 2302-9285, DOI: 10.11591/eei.v12i3.4633  1427

Noise estimation using an artificial neural network in the urban

area of Jaen, Cajamarca

Wendy Díaz1, Anali Tarrillo1, Candy Ocaña1,2, Lenin Quiñones1,2

1
Escuela Académica Profesional de Ingeniería Forestal y Ambiental, Universidad Nacional de Jaén, Perú
2
Instituto de Ciencia de Datos, Universidad Nacional de Jaén, Perú

Article Info ABSTRACT

Article history: Jaen is a city in constant urban growth which generates an increase in
vehicular traffic and active noise pollution. The research presents the
Received Aug 25, 2022 development of an artificial neural network (ANN) to estimate the noise
Revised Nov 2, 2022 produced by vehicular traffic in the urban area of the city. Consequently,
Accepted Nov 27, 2022 information was collected from two investigations coded as T1 and T2, for
which a matrix of 10 variables was elaborated with 210 and 273 data
respectively. Random random sampling was performed to divide the data
Keywords: matrix into 80% (training) and 20% (validation). Weka software and the
multi-layer perceptron (MLP) training algorithm were used to model the
Artificial intelligence ANN. An ANN for T1 with 6-19-1 architecture and an ANN for T2 with 6-
Expert system 15-1 architecture were obtained. The performance of the ANNs was evaluated
Information technology using the correlation coefficient (R), coefficient of determination (R2) and
Noise pollution root mean square error (RMSE). The results show that the MLP networks are
Urban traffic able to estimate the sound pressure level with values of
R=0.9927, R2=0.9854 and RMSE=0.7313 for T1, R=0.9989, R2=0.9978, and
RMSE=0.1515 for T2.
This is an open access article under the CC BY-SA license.

Corresponding Author:
Candy Ocaña
Escuela Académica Profesional de Ingeniería Forestal y Ambiental
Instituto de Ciencia de Datos, Universidad Nacional de Jaén
Calle Higinio Ortiz # 511, Jaén, Cajamarca, Perú
Email: [email protected]

1. INTRODUCTION
Noise pollution generates excessive noise that causes annoyance to living organisms. This noise is
considered one of the critical environmental problems because it endangers the health of the population;
causing affectations such as cardiovascular diseases, hearing problems, sleep disorders and adverse social
behaviour [1], [2], in addition to this, [3] found that this type of pollution influences the incidence and severity
of COVID-19, because it generates high levels of cortisol, weakening the immune system. Vehicle traffic is
one of the main sources of this pollution in cities [4], [5]. In general, there are two sets of factors that influence
noise annoyance: i) related to the physical characteristics of the sound (type of noise, level, duration, and
frequency spectrum), the time of day it occurs and the exposure and ii) related to the individual, including
physiological, psychological, and social characteristics that affect the subjective perception of noise [5], [6].
Vehicles are predominantly sources of low and medium frequency noise, which has a high penetrating
power and propagates with low dissipative absorption over long distances. The continuous growth of the car
fleet has progressively increased the need for special attention to urban traffic noise, which not only increases
in line with the growth of residential, industrial and commercial areas, but also causes adverse impact of noise
emissions on people [7], [8]. Predicting the level of noise produced by urban transport is an essential aspect of
mitigating environmental pollution. Therefore, it is necessary to have appropriate and specific mathematical

Journal homepage: http://beei.org

1428  ISSN: 2302-9285

tools (models) that can reproduce or simulate different acoustic scenarios for use in assessing and planning
urban planning activities [9], [10]. Artificial neural networks (ANNs) have recently emerged as an important
area of research, not only for their general ability to process noise data, but also to learn and store it, specifically
this computational model allows prediction and optimisation of traffic noise descriptors [4], [11], [12].
Studies have proposed method for assessing and predicting noise in the environment. These prediction
methods are mainly classified into three groups: physical propagation models, traditional statistical methods
and machine learning methods [13], [14]. Deep learning derives from the study of ANNs, which in many areas
of data science have demonstrated a remarkable ability to learn complex, non-linear relationships between sets
of variables [15], [16]. The ANN is inspired by the nature of real dynamic systems emulating the human brain,
where neurons are layered and interconnected by mathematical functions. Each neuron receives a weighted
signal from the previous layer, which is processed to learn from the examples provided by training algorithms
[17]. Common neural network models include multilayer perceptron (MLP), convolutional neural networks
(CNN), and recurrent neural networks (RNN) [18]. These algorithms iteratively update the model parameters
until the error between the actual value of the output variable and the experimental one is minimised
[4], [13], [19], [20].
The research proposes the use of a MLP ANN as a model for predicting urban environmental noise in
the city of Jaén. It was proposed to use the MLP network type for its versatility and prediction capacity. The
knowdlege discovery in databases (KDD) method was used for the development of the network, as a sequence
of ordered steps that allowed accurate information to be obtained. The modelling was carried out in Weka
software, introducing data obtained from environmental monitoring of research carried out in the same city,
with authors between 2016-2020; for training and validation we opted for the division of the data (80-20%) in
order to avoid the complexity of the model and therefore an overfitting of the model (overfitting).

2. METHOD
Research on sound evaluation in the urban area of Jaén for the period 2016-2020 was taken into
account. The databases were reviewed from institutional repositories of Peruvian universities and were
subjected to evaluation considering standardisation criteria in the information considered by the authors. We
opted for those data sources that consider the sound pressure level of the vehicle fleet and that consider the
variables (inclusion and exclusion criteria): name of road, location coordinates of sampling points, time and
date of data collection, maximum sound pressure level (Lmax), minimum sound pressure level (Lmin), number
of motokar per time unit and sampling point, number of linear moto per time unit and sampling point, number
of cars per time unit and sampling point, equivalent continuous sound pressure level (LAeqT). The ANNs were
developed using the KDD method, in the free software Weka and with the backpropagation learning algorithm
(80% of the data for training and 20% for validation).
To select independent variables influencing the equivalent continuous sound pressure level, the
CorrelationAtributeEval evaluator attribute was used, prioritising input variables with significance values
greater than or equal to 0.1. The correlation coefficient (R), the coefficient of determination (R 2) and the root
mean square error (RMSE) are selected as criteria for evaluating model performance. To visualise the
relationship between the actual and the protonistic sound pressure level (LAeqT), the variables used in Weka
were simulated in SPSS based on linear regression. Figure 1 (in Appendix) shows the methodological flow for
the development of the proposed neural network.

3. RESULTS AND DISCUSSION

3.1. Collection of noise pollution data from the urban area of Jaen
The institutional repositories of the country's universities were consulted during the period
2016-2020. Five data sources were preliminarily evaluated (Table 1), of which two met the inclusion and
exclusion criteria previously detailed: i) sound pressure level by the vehicle fleet in the city of Jaén, from
December 2018 to February 2019 (T1) and ii) assessment of vehicular noise pollution based on the supreme
decree N°085-2003-PCM regulation of environmental quality standards for noise carried out in the province
of Jaen, Department of Cajamarca, 2016 (T2). It was verified that the data collected and considered in the study
were obtained by sound level meters calibrated by the National Institute of Quality (INACAL) of Peru.

3.2. Artificial neural network for estimating noise pollution in the urban area of Jaen
3.2.1. Attribute selection for ANN, T1 and T2
The data obtained for both T1 and T2 were divided into 80% for training and 20% for validation. The
CorrelationAttributeEval attribute evaluator and the ranker search method in Weka software were used with
80% of the data for both T1 and T2, obtaining the importance values for each input variable.

Bulletin of Electr Eng & Inf, Vol. 12, No. 3, June 2023: 1427-1434
Bulletin of Electr Eng & Inf ISSN: 2302-9285  1429

Tables 2 and 3 show the nine importance values for each variable evaluated for T1 and T2, however, in order
to propose the number of input variables, the values of importance <0.1 were chosen, reducing the input
variables for the models to six.

Table 1. Sources of literature review

Name of research Author University
Evaluation of sound pressure levels in commercial establishments in the Gianela Olivera Zurita National
urban area of the city of Jaen, based on supreme decree N°085-2003-PCM Kiara Belkiss Silva Vega University of Jaen
Environmental Quality Standards (ECAS) for noise in the main university Felipe Nery Silva Cabrera National
higher education centres in the city of Jaen University of Jaen
Sound pressure levels in the markets of the city of Jaen, Cajamarca -2019 Katiri Tatiana Estela Carranza National
Jefferson Jair Goicochea Pérez University of Jaen
Evaluation of vehicular noise pollution based on the Supreme Decree N°085- Cintia Karely Cruzado National
2003-PCM Regulation of Environmental Quality Standards for Noise carried Ancajima University of Jaen
out in the province of Jaén, department of Cajamarca, 2016 Yanira Susana Soto Medina
Sound pressure level by the vehicle fleet in the city of Jaén, December 2018 Elser Burga Mendoza National
to February 2019 University of Jaen

Table 2. T1, 80% attribute selection

Variable Importance values
Lmax 0.981
Lmin 0.7142
CoordenadaUTM 0.2799
NombredelaVía 0.2799
Hora 0.155
Motokar 0.1481
Carros 0.0868
MotoLineal 0.0536
Fecha -0.0327

Table 3. T2, 80%, attribute selection

Variable Importance values
Lmax 0.8943
Lmin 0.3823
MotoLineal 0.1882
Motokar 0.1818
CoordenadasUTM 0.1543
NombredelaVía 0.1475
Carros 0.0788
Hora 0.0655
Fecha 0.051

3.2.2. Training of ANNs

For T1, a 6-19-1 architecture ANN was obtained; six (6) input neurons, a hidden layer with nineteen
(19) nodes and an output layer with the dependent variable LAeqT, Figure 2(a); on the contrary, for T2, a 6-
15-1 architecture ANN was obtained; six (6) input neurons, a hidden layer with fifteen (15) nodes and an output
layer with the dependent variable LAeqT, Figure 2(b). The fit statistics used for the training set of ANN-T1
and ANN-T2 were R, R2 and RMSE (Table 4); acceptable values were obtained for both models. For T1 and
T2 the R and R2 presented values close to unity showing a good performance of the model; unlike the values
obtained for the RMSE, which showed values close to unity and therefore a higher error rate, this due to the
rate of learning in the training process (high learning rates will make the training converge faster, but the fit of
the trainable parameters will be less accurate and will result in higher error rates) [21]. The results of the
training of the networks are shown in Figures 3 and 4, where the results of the dispersion of predicted LAeqt
values and actual LAeqt are graphically represented. The relationship between the observed LAeqt and the
estimated LAeqt, for model T1, presents a line of fit of equation
y=2.48+1.01 x, with coefficient of determination R2=0.98, on the contrary, model T2 presents a line of fit of
equation y=13.05+0.83 x, with R2=0.92.

3.3. Validation of the artificial neural network

For the validation of the ANNs, the same adjustment statistics used in the training stage were used, as
shown in Table 5. However, in this case 20% of the data from each investigation was used. For both models’

Noise estimation using an artificial neural network in the urban area of Jaen, Cajamarca (Wendy Díaz)
1430  ISSN: 2302-9285

good values of R and R2 were obtained due to their closeness to unity; differing in the RMSE, as T2 presented
a more positive value compared to T1, but for both models’ acceptable values were obtained.

(a)

(b)

Figure 2. Validation of the ANNs (a) architecture of T1-TNA and (b) architecture of T2-TNNA

Table 4. Values of statistics used in RNA-T1 and RNA-T2 training

Statistic RNA-T1 RNA-T2
R 0.9861 0.9606
R2 0.9723 0.9227
RMSE 0.7952 0.99

Bulletin of Electr Eng & Inf, Vol. 12, No. 3, June 2023: 1427-1434
Bulletin of Electr Eng & Inf ISSN: 2302-9285  1431

Figure 3. Predicted LAeqT vs. actual LAeqT of T1

Figure 4. Predicted LAeqT vs. actual LAeqT of T2

Table 5. Values of statistics used in validation of RNA-T1 and RNA-T2

Statistic RNA-T1 RNA-T2
R 0.9927 0.9989
R2 0.9854 0.9978
RMSE 0.7313 0.1515

The performance of the neural network was obtained through the R, R 2 and RMSE statistics. The
networks obtained show good performance in both the training and validation stages, giving values close to
unity, which demonstrates a positive relationship between the data obtained by the network and the data
provided. It should be noted that the validation values were more significant; obtaining R=0.9927 and
R2=0.9854, RMSE=0.7313 at T1 and R=0.9989 and R2=0.9978, RMSE=0.1515 at T2. If these results are
compared with those obtained by different authors such as Mansourkhaki et al. [22] who obtained an R=0.992
and R2=0.983, RMSE=0.1515 at T2 and the case of Sequeira et al. [23] who obtained
R=0.995, R2=0.991, and RMSE=0.44; it can be said that the networks obtained for both T1 and T2 show high
efficiency for noise prediction. It can also be affirmed that MLP ANNs for both T1 and T2 show high efficiency
for the prediction of the equivalent continuous sound pressure level (LAeq); these results are analogous to
Noise estimation using an artificial neural network in the urban area of Jaen, Cajamarca (Wendy Díaz)
1432  ISSN: 2302-9285

those obtained by Genaro García where he compared the efficiency of an ANN with mathematical models
(RSLS 90 and technical guidelines for noise impact assessment (Criterion)). Obtaining that the ANN has a
better performance and by Chen et al. [24] comparing two types of neural networks, including the MLP
network and the radial basis function (RBF) network for predicting traffic noise, where it was shown that the
MLP network performed better than the RBF network in predicting noise level [25].

4. CONCLUSION
An ANN model was developed for the estimation of the equivalent continuous sound pressure level
(LAeqT) using the MLP etwork type and algorithm, using the variables that contributed most to the estimation
of the dependent variable during the training stage. For T1-Burga Mendoza they were: name of road, UTM
coordinates, time, Lmax, Lmin, and motokar, for T2-Cruzado Ancajima and Soto Medina they were: name of
road, UTM coordinates, Lmax, Lmin, Motokar, and Moto linear. A structure of 6-19-1 was obtained for T1-
Burga Mendoza and 6-15-1 for T2-Cruzado Ancajima and Soto Medina. The validation results of the work
show that the network created for T1 is capable of estimating the sound pressure level with R=0.9927 and
R2=0.9854 and for T2 with R=0.9989 and R2=0.9978.

ACKNOWLEDGEMENTS
The authors would like to thank the Universidad Nacional de Jaén for the facilities to carry out the
research. To the professionals Elser Burga Mendoza, Yanira Soto Medina and Cintia Cruzado Ancajima for
providing us with their research data.

APPENDIX

Figure 1. Method

Bulletin of Electr Eng & Inf, Vol. 12, No. 3, June 2023: 1427-1434
Bulletin of Electr Eng & Inf ISSN: 2302-9285  1433

REFERENCES
[1] S. V. Kuznetsov et al., “Emotional artificial neural network (EANN)-based prediction model of maximum A-weighted noise
pressure level,” Noise Mapping, vol. 9, no. 1, pp. 1–9, Jan. 2022, doi: 10.1515/noise-2022-0001.
[2] P. E. Baffoe, A. A. Duker, and E. V. S-Kwarteng, “Assessment of health impacts of noise pollution in the Tarkwa Mining Community of
Ghana using noise mapping techniques,” Global Health Journal, vol. 6, no. 1, pp. 19–29, Mar. 2022, doi: 10.1016/j.glohj.2022.01.005.
[3] J. Díaz, J. A-López-Bueno, D. Culqui, C. Asensio, G. S-Martínez, and C. Linares, “Does exposure to noise pollution influence the
incidence and severity of COVID-19?,” Environ Res, vol. 195, p. 110766, Apr. 2021, doi: 10.1016/j.envres.2021.110766.
[4] V. Nourani, H. Gökçekuş, I. K. Umar, and H. Najafi, “An emotional artificial neural network for prediction of vehicular traffic
noise,” Science of The Total Environment, vol. 707, p. 136134, Mar. 2020, doi: 10.1016/j.scitotenv.2019.136134.
[5] L. B-Moncayo, J. L-Naranjo, M. Chávez, I. P-García, and C. Garzón, “A machine learning approach for traffic-noise annoyance
assessment,” Applied Acoustics, vol. 156, pp. 262–270, Dec. 2019, doi: 10.1016/j.apacoust.2019.07.010.
[6] J. Tomić, N. Bogojević, and D. Š. Pavlović, “Application of Soft Computing Techniques in Prediction of Road Traffic Noise
Levels,” in 2018 26th Telecommunications Forum (℡FOR), Nov. 2018, pp. 1–7, doi: 10.1109/℡FOR.2018.8611931.
[7] I. Deryabin, “Traffic lights as a factor in the regulation of traffic noise,” Transportation Research Procedia, vol. 63, pp. 436–440,
Jan. 2022, doi: 10.1016/j.trpro.2022.06.032.
[8] C. Guarnaccia, J. Quartieri, and C. Tepedino, “A hybrid predictive model for acoustic noise in urban areas based on time series
analysis and artificial neural network,” AIP Conference Proceedings, vol. 1836, no. 1, p. 020069, Jun. 2017, doi: 10.1063/1.4982009.
[9] F. Cirianni and G. Leonardi, “The application of a neural network on a study of noise pollution in urban transport: a case in Villa
S. Giovanni,” Jan. 2004, pp. 559–570.
[10] A. S. R. Subramanian, T. Gundersen, and T. A. Adams, “Modeling and Simulation of Energy Systems: A Review,” Processes, vol.
6, no. 12, Dec. 2018, doi: 10.3390/pr6120238.
[11] L. J. R. Nunes, A. M. Rodrigues, L. M. E. F. Loureiro, L. C. R. Sá, and J. C. O. Matias, “Energy Recovery from Invasive Species:
Creation of Value Chains to Promote Control and Eradication,” Recycling, vol. 6, no. 1, Mar. 2021, doi: 10.3390/recycling6010021.
[12] Kumar, “Traffic noise prediction and optimization using response surface method (RSM),” Arabian Journal of Geosciences, pp. 1–
16, Oct. 2021, doi: 10.1007/s12517-021-08547-5.
[13] X. Zhang, M. Zhao, and R. Dong, “Time-Series Prediction of Environmental Noise for Urban IoT Based on Long Short-Term
Memory Recurrent Neural Network,” Applied Sciences, vol. 10, no. 3 Jan. 2020, doi: 10.3390/app10031144.
[14] U. Masood, H. Farooq, and A. Imran, “A Machine Learning Based 3D Propagation Model for Intelligent Future Cellular Networks,”
in 2019 IEEE Global Communications Conference (GLOBECOM), Dec. 2019, pp. 1–6, doi:
10.1109/GLOBECOM38437.2019.9014187.
[15] A. S. Tenney, M. N. Glauser, C. J. Ruscher, and Z. P. Berger, “Application of Artificial Neural Networks to Stochastic Estimation
and Jet Noise Modeling,” AIAA Journal, vol. 58, no. 2, pp. 647–658, 2020, doi: 10.2514/1.J058638.
[16] E. Turajlic, A. Begović, and N. Škaljo, “Application of Artificial Neural Network for Image Noise Level Estimation in the SVD
domain,” Electronics, vol. 8, no. 2, Feb. 2019, doi: 10.3390/electronics8020163.
[17] V. K. Mishra, K. Anand, and A. Bhardwaj, “Clustering assisted artificial neural network for handling noisy big data: An application
for estimation of parameter in combined mode conduction and radiation heat transfer,” Heat Transfer, vol. 51, no. 6, pp. 5386–
5416, 2022, doi: 10.1002/htj.22552.
[18] R. Vinayakumar, M. Alazab, K. P. Soman, P. Poornachandran, A. Al-Nemrat, and S. Venkatraman, “Deep Learning Approach for
Intelligent Intrusion Detection System,” IEEE Access, vol. 7, pp. 41525–41550, 2019, doi: 10.1109/ACCESS.2019.2895334.
[19] R. M-Medina, R. P-Gálvez, A. Guadix, and E. M. Guadix, “Artificial neuronal network modeling of the enzymatic hydrolysis of
horse mackerel protein using protease mixtures,” Biochemical Engineering Journal, vol. 105, pp. 364–370, Jan. 2016, doi:
10.1016/j.bej.2015.10.009.
[20] Z. Zhang, X. Deng, and L. Zheng, “A review on varying-parameter convergence differential neural network,” Neurocomputing,
vol. 490, pp. 54–65, Jun. 2022, doi: 10.1016/j.neucom.2022.03.026.
[21] J. M-Oreja and J. Gozalvez, “Predicción de la velocidad del tráfico basada en redes neuronales convolucionales,” Avances en
arquitectura y tecnología de computadores. Actas de las Jornadas SARTECO 2017, pp. 677–684, Sep. 2017, doi:
10.5281/zenodo.996051.
[22] A. Mansourkhaki, M. Berangi, M. Haghiri, and M. Haghani, “A neural network noise prediction model for Tehran urban roads,”
Journal of Environmental Engineering and Landscape Management, vol. 26, no. 2, Jun. 2018, doi:
10.3846/16486897.2017.1356327.
[23] M. E. Sequeira, V. H. Cortínez, and A. P. Azzurro, “Caracterización Acústica de Ambientes Urbanos Mediante el Uso de Redes
Neuronales,” Mecánica Computacional, vol. 37, no. 5, Nov. 2019.
[24] L. Chen, T. Liu, B. Tang, H. Xiang, and Q. Sheng, “Modelling traffic noise in a wide gradient interval using artificial neural
networks,” Environmental Technology, vol. 42, no. 23, pp. 3561–3571, Oct. 2021, doi: 10.1080/09593330.2020.1734098.
[25] A. Mansourkhaki, M. Berangi, and M. Haghiri, “Comparative Application of Radial Basis Function and Multilayer Perceptron
Neural Networks to Predict Traffic Noise Pollution in Tehran Roads,” J. Ecol. Eng., vol. 19, no. 1, pp. 113–121, Jan. 2018, doi:
10.12911/22998993/79411.

BIOGRAPHIES OF AUTHORS

Wendy Díaz she received his degree in Forestry and Environmental Engineering
from the National University of Jaen (UNJ) in Peru in 2021 and is now a Master's student in
Environmental Management at the National University of Trujillo (UNT) in Peru. She can
be contacted at email: [email protected].

Noise estimation using an artificial neural network in the urban area of Jaen, Cajamarca (Wendy Díaz)
 1434  ISSN: 2302-9285

Anali Tarrillo received the degree of Forestry and Environmental Engineer

from the Universidad Nacional de Jaén in 2021. Currently, she works as an engineer in the
private company ZURSAN INGENIERIA Y CONSTRUCCIÓN EIRL in the area of Safety,
Occupational Health and Environment. Her current research interests include artificial
intelligence and deep learning. She can be contacted at email: [email protected].

Candy Ocaña formed by the Faculty of Agricultural Sciences of the National

University of Cajamarca, profession forest engineer, master's degree in public management
and completed studies of the master's degree in Forestry and Natural Resource Management,
currently a doctoral student in Economics of Natural Resources. With knowledge and
application of research methodologies for the elaboration and execution of scientific
projects, administration of research projects and public investment. Work and professional
experience in undergraduate university teaching at the University of Cajamarca, currently
teaching in the Assistant category at the National University of Jaen (UNJ). She can be
contacted at email: [email protected].

Lenin Quiñones Bachelor's degree in Mathematics, Master's degree in

Computer Science and Ph.D in Industrial Engineering. Work and professional experience in
undergraduate university teaching in Chile and Peru and postgraduate in the Master of
Science with mention in Environmental Management of the Faculty of Ecology of the
Universidad Nacional de San Martín. Currently teaching undergraduate and graduate courses
at the National University of Jaen. The lines of research that I develop: Construction and use
of mathematical models for Engineering, application of Artificial Intelligence, computation
as a tool for solving research problems and teaching of Mathematics. Organizer and speaker
at national and international conferences, thesis advisor, reviewer of scientific journals,
author of scientific articles and books. He can be contacted at email:
[email protected].

Bulletin of Electr Eng & Inf, Vol. 12, No. 3, June 2023: 1427-1434

View publication stats