TY - JOUR
T1 - Provincial waste generation prediction using evolutionary-based neuro-fuzzy model
T2 - International Conference on Engineering for a Sustainable World, ICESW 2022
AU - Adeleke, Oluwatobi
AU - Jen, Tien Chien
N1 - Publisher Copyright:
© 2023 Elsevier Ltd. All rights reserved. Selection and peer-review under responsibility of the scientific committee of the International Conference on Engineering for a Sustainable World.
PY - 2022
Y1 - 2022
N2 - The accurate prediction of waste generation at the provincial level is essential for effective waste management planning in South Africa. However, the complex and uncertain nature of waste generation patterns poses significant challenges for policymakers and waste management authorities. Existing prediction models often overlook the intricate relationships between various factors influencing waste generation. Therefore, there is a need to develop an advanced predictive model that integrates evolutionary algorithms and neuro-fuzzy systems to provide reliable and comprehensive waste generation predictions. In this study, adaptive neuro-fuzzy inference system (ANFIS) model with two evolutionary algorithms namely particle swarm optimization (PSO) and genetic algorithm (GA) were developed to predict the provincial waste quantities generated using South Africa as a case study. Relevant demographic and economic variables of nine provinces in South Africa namely population, provincial GDP, household number and employment data were set of input variables while the waste quantity generated was set as output variable. The fuzzy c-means (FCM) clustering technique developed the fuzzy inference system based on its speed boost capability. Hyper-parameters setting of the FCM-clustered neuro-fuzzy model in a range of 2-7 clusters was tested. The models' accuracy was evaluated using known statistical metrics like root mean square error (RMSE), Average Absolute Percentage Relative Error (AAPRE), mean absolute deviation (MAD), and co-efficient of determination (R2). The most accurate model is the FCM-ANFIS-GA with 5-clusters giving RMSE, AAPRE, MAD and R2-values of 0.7335, 0.3656, 7.3359 and 0.9986 at the training phase and 0.8041, 0.3291, 8.1467 and 0.9962 at the testing phase. The developed model would assist in sustainable waste management and would be beneficial to South African Waste Information Center (SAWIC) and other countries.
AB - The accurate prediction of waste generation at the provincial level is essential for effective waste management planning in South Africa. However, the complex and uncertain nature of waste generation patterns poses significant challenges for policymakers and waste management authorities. Existing prediction models often overlook the intricate relationships between various factors influencing waste generation. Therefore, there is a need to develop an advanced predictive model that integrates evolutionary algorithms and neuro-fuzzy systems to provide reliable and comprehensive waste generation predictions. In this study, adaptive neuro-fuzzy inference system (ANFIS) model with two evolutionary algorithms namely particle swarm optimization (PSO) and genetic algorithm (GA) were developed to predict the provincial waste quantities generated using South Africa as a case study. Relevant demographic and economic variables of nine provinces in South Africa namely population, provincial GDP, household number and employment data were set of input variables while the waste quantity generated was set as output variable. The fuzzy c-means (FCM) clustering technique developed the fuzzy inference system based on its speed boost capability. Hyper-parameters setting of the FCM-clustered neuro-fuzzy model in a range of 2-7 clusters was tested. The models' accuracy was evaluated using known statistical metrics like root mean square error (RMSE), Average Absolute Percentage Relative Error (AAPRE), mean absolute deviation (MAD), and co-efficient of determination (R2). The most accurate model is the FCM-ANFIS-GA with 5-clusters giving RMSE, AAPRE, MAD and R2-values of 0.7335, 0.3656, 7.3359 and 0.9986 at the training phase and 0.8041, 0.3291, 8.1467 and 0.9962 at the testing phase. The developed model would assist in sustainable waste management and would be beneficial to South African Waste Information Center (SAWIC) and other countries.
KW - ANFIS
KW - Evolutionary algorithms
KW - GA
KW - PSO
KW - South Africa
KW - Waste quantities
UR - http://www.scopus.com/inward/record.url?scp=85196123562&partnerID=8YFLogxK
U2 - 10.1016/j.matpr.2023.06.226
DO - 10.1016/j.matpr.2023.06.226
M3 - Conference article
AN - SCOPUS:85196123562
SN - 2214-7853
VL - 105
SP - 37
EP - 43
JO - Materials Today: Proceedings
JF - Materials Today: Proceedings
IS - C
Y2 - 19 May 2022 through 20 May 2022
ER -
