LOGO Deep Learning Neural Networks with Metaheuristic Optimization

  • Slides: 15
Download presentation
LOGO Deep Learning Neural Networks with Metaheuristic Optimization Algorithms for Groundwater Contamination Vulnerability Mapping

LOGO Deep Learning Neural Networks with Metaheuristic Optimization Algorithms for Groundwater Contamination Vulnerability Mapping in Miryang Aquifer, South Korea Hussam Eldin Elzain · Sang Yong Chung Venkatramanan Senapathi · Kye-Hun Park Pukyong National University, Korea

6/4/2021 Research Scope v Assessment of groundwater contamination vulnerability v Using Deep learning neural

6/4/2021 Research Scope v Assessment of groundwater contamination vulnerability v Using Deep learning neural networks: Artificial Neural Fuzzy Logic Inference System (ANFIS) v Combined with Genetic Algorithm (GA) and Particle Swarm Optimization (PSO) v Input data: DRASTIC 7 Factors + Landuse v Target (output) data: 95 NO 3 -N concentrations v Study area: Miryang City, Korea (South) 2 Groundwater Environment Lab. , PKNU, Korea

6/4/2021 Study Area (a) (b) (c) Fig. 1. (a) Location map, (b) Sample points,

6/4/2021 Study Area (a) (b) (c) Fig. 1. (a) Location map, (b) Sample points, (c) Geological map. 3 Groundwater Environment Lab. , PKNU, Korea

6/4/2021 Flow Chart of VI Assessment 4 Groundwater Environment Lab. , PKNU, Korea

6/4/2021 Flow Chart of VI Assessment 4 Groundwater Environment Lab. , PKNU, Korea

6/4/2021 Artificial Neural Fuzzy Logic Inference System (ANFIS) 5 Groundwater Environment Lab. , PKNU,

6/4/2021 Artificial Neural Fuzzy Logic Inference System (ANFIS) 5 Groundwater Environment Lab. , PKNU, Korea

6/4/2021 Evaluation of vulnerability factors using Pearson technique and collinearity statistics DRASTIC Factors Predictive

6/4/2021 Evaluation of vulnerability factors using Pearson technique and collinearity statistics DRASTIC Factors Predictive Power Depth to water Collinearity Statistics Tolerance VIF 0. 551 0. 512 1. 951 Recharge 0. 454 0. 574 1. 739 Aquifer media 0. 566 0. 374 2. 672 Soil media 0. 460 0. 534 1. 872 Topographic slope 0. 338 0. 619 1. 389 Vadose zone 0. 595 0. 376 2. 657 Conductivity 0. 749 0. 506 1. 975 Land use 0. 671 0. 584 1. 710 6 Groundwater Environment Lab. , PKNU, Korea

6/4/2021 Thematic Maps of DRASTIC Factors (a) Depth to water (b) Recharge rate (c)

6/4/2021 Thematic Maps of DRASTIC Factors (a) Depth to water (b) Recharge rate (c) Aquifer media (d) Soil media (a) (b) (d) (c) 7 Groundwater Environment Lab. , PKNU, Korea

6/4/2021 Thematic Maps of DRASTIC Factors (e)Topographic slope (f)Impact to vadose zone (g)Hydraulic Conductivity(h)Landuse

6/4/2021 Thematic Maps of DRASTIC Factors (e)Topographic slope (f)Impact to vadose zone (g)Hydraulic Conductivity(h)Landuse (e) (e) (g) (f) (h) 8 Groundwater Environment Lab. , PKNU, Korea

6/4/2021 Graphical representations of training and testing performances for (a, d) Original ANFIS, (b,

6/4/2021 Graphical representations of training and testing performances for (a, d) Original ANFIS, (b, e) ANFIS-GA, and (c, f) ANFIS-PSO. (a) (d) (b) (c) (e) (f) 9 Groundwater Environment Lab. , PKNU, Korea

6/4/2021 Statistical performance of ANFIS models Statistical Factors Original ANFIS-GA ANFIS-PSO Training Testing MAE

6/4/2021 Statistical performance of ANFIS models Statistical Factors Original ANFIS-GA ANFIS-PSO Training Testing MAE 0. 190 0. 637 0. 103 0. 333 0. 100 0. 318 RMSE 0. 338 0. 890 0. 319 0. 491 0. 234 0. 478 0. 931 0. 493 0. 935 0. 717 0. 971 0. 851 10 Groundwater Environment Lab. , PKNU, Korea

6/4/2021 Validation of Models (a) AUC validation curves of ANFIS models for (a) training

6/4/2021 Validation of Models (a) AUC validation curves of ANFIS models for (a) training and (b) testing results (b) Variation curves of objective function values produced from ANFIS-MOA models. 11 Groundwater Environment Lab. , PKNU, Korea

6/4/2021 Groundwater contamination (nitrate) vulnerability maps (a) Original NO 3 -N, (b) Original ANFIS,

6/4/2021 Groundwater contamination (nitrate) vulnerability maps (a) Original NO 3 -N, (b) Original ANFIS, (c) ANFIS-GA, (d) ANFIS-PSO. (a) (c) (b) (d) 12 Groundwater Environment Lab. , PKNU, Korea

6/4/2021 Conclusions (1) v This study used adaptive neuro-fuzzy inference system (ANFIS) combined with

6/4/2021 Conclusions (1) v This study used adaptive neuro-fuzzy inference system (ANFIS) combined with metaheuristic optimization algorithms (MOA) of genetic algorithm (GA) and particle swarm optimization (PSO) to assess groundwater contamination (NO 3 -N) vulnerability in Miryang, City, South Korea. v 8 parameters (Depth to water, net Recharge, Aquifer type, Soil media, Topographic slope, Impact to vadose zone, hydraulic Conductivity and Landuse) were classified into numerical ratings to be used as input variables of ANFIS-MOA models. v ANFIS, ANFIS-PSO and ANFIS-GA used 95 measured NO 3 N concentrations from monitoring wells for the training (70% data) and testing (30% data) purposes. 13 Groundwater Environment Lab. , PKNU, Korea

6/4/2021 Conclusions (2) v The model performance was verified by the statistical data of

6/4/2021 Conclusions (2) v The model performance was verified by the statistical data of MAE, RMSE and correlation coefficient, and it was also verified by ROC/AUC curve. ANFIS-PSO showed the highest performance in the testing results (MAE= 0. 318; RMSE=0. 478; R=0. 851; AUC=0. 893) in comparison with ANFIS-GA (MAE=0. 333; RMSE=0. 491; R=0. 717; AUC=0. 821) and Original ANFIS (MAE=0. 637; RMSE=0. 890; R=0493; AUC=0. 643). v Integrated studies of statistical errors, objective function values, ROC/AUC curve and groundwater contamination (NO 3 -N) vulnerability maps revealed that ANFIS-PSO model was the most superior in predicting groundwater contamination vulnerability in Miryang City of South Korea. 14 Groundwater Environment Lab. , PKNU, Korea

LOGO

LOGO

LOGO Wafa Setia LOGO LOGO LOGO LOGO LOGOLOGO Wafa Setia LOGO LOGO LOGO LOGO LOGO
Neural networks 1 Neural networks Neural networks areNeural networks 1 Neural networks Neural networks are
Neural networks 1 Neural networks Neural networks areNeural networks 1 Neural networks Neural networks are
Neural networks 1 Neural networks Neural networks areNeural networks 1 Neural networks Neural networks are
Deep Neural Network optimization Binary Neural Networks BarryDeep Neural Network optimization Binary Neural Networks Barry
Finger Biometric Neural networks 1 Neural networks NeuralFinger Biometric Neural networks 1 Neural networks Neural
Deep neural networks DNNs Conventional and deep networksDeep neural networks DNNs Conventional and deep networks
A New Metaheuristic for Optimization Optics Inspired OptimizationA New Metaheuristic for Optimization Optics Inspired Optimization
Stochastic Neural Networks Deep Learning and Neural NetsStochastic Neural Networks Deep Learning and Neural Nets
CSCI 5922 Neural Networks and Deep Learning DeepCSCI 5922 Neural Networks and Deep Learning Deep
CSCI 5922 Neural Networks and Deep Learning DeepCSCI 5922 Neural Networks and Deep Learning Deep
Artificial Neural Networks 1 Neural networks Networks ofArtificial Neural Networks 1 Neural networks Networks of
Artificial Neural Networks Artificial Neural Networks Interconnected networksArtificial Neural Networks Artificial Neural Networks Interconnected networks
Artificial Neural Networks Artificial Neural Networks Interconnected networksArtificial Neural Networks Artificial Neural Networks Interconnected networks