Machine learning aided multi-objective optimization and multi-criteria decision making: Framework and two applications in chemical engineering

AbstractIn machine learning, classification is one of the important methods used to train a model for the identification of each class on a labelled dataset. The performance of the algorithm can be enhanced using ensemble techniques such as bagging and random subsampling. Random forest (RF) is an ensemble machine learning algorithm to address multiclass classification problems. Several RF algorithms have been proposed to obtain various levels of accuracy on different datasets during classification. But it has been observed that the manual configuration of RF hyperparameter and its accuracy varies significantly depending upon different datasets. Manual configuration of hyperparameter is a time consuming job. Thus, the hyperparameter optimization process is used to generate optimal hyperparameter efficiently. Hyperparameter optimization also enhances the performance of machine learning algorithms. This paper proposes improved particle swarm optimization (PSO) method as hyperparameter optimizer to identify the optimal hyperparameter of the random forest algorithm. We compare the performance of RF model using improved PSO optimizer and existing RF model with default hyperparameter. The improved PSO outperforms the improvement of accuracy during the classification of each digit of the handwritten dataset.KeywordsRandom forestPSOHyperparameterOptimization 

Optimization of Random Forest Hyperparameter Using Improved PSO for Handwritten Digits Classification

The production of solar energy at altitudes higher than 3800 m above sea level is not constant because the relevant factors are highly varied and complex due to extreme solar radiation, climatic variations, and hostile environments. Therefore, it is necessary to create efficient prediction models to forecast solar production even before implementing photovoltaic systems. In this study, stacking techniques using ElasticNet and XGBoost were applied in order to develop regression models that could collect a maximum number of features, using the LASSO, Ridge, ElasticNet, and Bayesian models as a base. A sequential feature selector (SFS) was used to reduce the computational cost and optimize the algorithm. The models were implemented with data from a string photovoltaic (PV) system in Puno, Peru, during April and August 2021, using 15 atmospheric and photovoltaic system variables in accordance with the European standard IEC 61724-20170. The results indicate that ElasticNet reduced the MAE by 30.15% compared to the base model, and that the XGBoost error was reduced by 30.16% using hyperparameter optimization through modified random forest research. It is concluded that the proposed models reduce the error of the prediction system, especially the stacking model using XGBoost with hyperparameter optimization.

/pdf/multiparameter-regression-of-a-photovoltaic-system-by-26ol8l5y.pdf

Multiparameter Regression of a Photovoltaic System by Applying Hybrid Methods with Variable Selection and Stacking Ensembles under Extreme Conditions of Altitudes Higher than 3800 Meters above Sea Level

In computational neuroscience, the automatic fitting of neuron parameters of simple neurons based on data or other detailed models is a challenging problem. Stochastic variability of firing activity in neurons makes an automated fitting, a complex neuronal dynamic problem. In this study, four features of neuronal firing including spike amplitude, spike count, spike timing, and inter-spike interval were employed to create a simplified neuronal network model using particle swarm optimization and genetic algorithm. This paper showcases a methodology to reduce detailed models that arise from experimental data into simple spiking Izhikevich models, which can be computationally more effective while reconstructing large-scale circuits for behavioral modeling. The approach is illustrated during spontaneous firing activity and current injection to validate the model. The result indicated that the reduced neuronal model shows matching firing activity when optimize with optimization algorithms. The four features of the neuronal activity were matched with the experimental data after optimization. The study also analyzed the parameter fitting accuracy and runtime efficiency of reduction based on the optimization algorithms. The result indicated that the reduction based on particle swarm optimization showed less error percentage while reducing the model when compared with the genetic algorithm. 

Automated Reduction of Detailed Biophysical Cerebellar Neurons to Izhikevich Neurons

This paper develops an optimization framework based on evolutionary computation for the multi-step prediction enhancement of Deep Learning-based Dynamic Systems simulation models. We propose using the Differential Evolution algorithm and an Autoencoder network to find the optimal arrangement that accurately models a nonlinear system. A series of experiments are performed using a nonlinear, chaotic benchmark system: the double pendulum to validate our claims. As a result, the prediction error and confidence level were increased by an average of 20% against conventional parameters selection methods. Furthermore, we found that the training stage relies less on trial and error approaches in favor of a quantitative objective function using an optimization method. 

Evolutionary Optimization of Multi-step Dynamic Systems Learning

\n In this study, genetic algorithms (GAs) and particle swarm optimization (PSO) are used to make an automated choice of hyperparameters of the multilayer perceptron (MLP)-NARX, extreme learning machine (ELM)-NARX, and echo state network (ESN)-NARX neural models applied to the identification of two photovoltaic systems: one installed in Teresina, in Brazil, and another in Hamburg, Germany. The automatic optimization process results showed that the PSO algorithm presents superior performance compared to the GA algorithm. Likewise, the identification carried out aimed to estimate the power generated by photovoltaic systems from two different approaches: linear mathematical models and neural identification models. Thus, the neural models implemented are more efficient and accurate than the linear mathematical models compared. From accuracy, the neural models ESN-NARX and MLP-NARX were considered the best in identifying Hamburg and Teresina’s photovoltaic systems, respectively.

Optimization of NARX Neural Models Using Particle Swarm Optimization and Genetic Algorithms Applied to Identification of Photovoltaic Systems

Este trabalho tem como Finalidade apresentar o desenvolvimento de um sistema de telemetria automotiva capaz de capturar dados do sistema eletromecânico de um veiculo, para que estes possam descrever o comportamento do automovel e de seu condutor. Desta forma, este trabalho propoe construir um sistema de aquisicao de dados atraves do sistema de diagnostico OBD II, construir primitivas que permitam guardar essas informacoes remotamente e, por um, empregar algoritmos de reconhecimento de padroes, com o intuito de evidenciar modelos de conducao de um automovel, classificando os perfis de conducao dos motoristas para que seja possivel sugerir mudancas de comportamento e prevenir falhas mecânicas. Os resultados obtidos sugerem que os atributos utilizados na classificacao permitem uma taxa de acerto superior a 90 %, com destaque para a rede ELM em custo computacional

José Maria Pires de Menezes Júnior

Papers

Optimization of NARX Neural Models Using Particle Swarm Optimization and Genetic Algorithms Applied to Identification of Photovoltaic Systems

Perfil de condução de motoristas: análise por meio de algoritmos de reconhecimento de padrões