Optimal Number of Hidden Neuron Identification for Sustainable Manufacturing Application

Abstract

There were 50 data sample obtained from industries in Malaysia that practice sustainable manufacturing. Input file is presented in matrix 4x50 and 1x50 matrix as target file. However, there is no suitable number of hidden neuron that can be applied for the neural network model with 4 inputs and 1 output. An experiment has been done to identify the suitable hidden neuron through the observation of values from MSE and Regression. The hidden neuron must be no overfitting. The same goes for output and targets value must have close or linear relationship. The sample of tested hidden neuron is from 5 to 40 hidden neurons. The final answer obtained after look into Mean Square Error (MSE) values, Regression values and plots is hidden neuron 29. Hidden neuron 29 shows positive result in all criteria and should be implemented for this type of neural network model.

An Artificial Intelligence Approach for Estimating the Turbidity of Artisanal Wine and Dosage of Clarifying Agents

“…curve (Figure 2) also displays the efficiency of the neural network model observing a good fit between predicted and experimental values, as mentioned by Noor et al [30] and Sahin et al [38], which is also explained for the upper value of R 2 , justifying the selection. The metrics and the selected algorithm indicate that the model is quite robust because it updates the weight and bias values at the culmination of each epoch, as stated by Incio et al [39].…”

“…In addition, it also presented some of the lowest values for MSE, RNSME, and AIC, demonstrating the feasibility of the supervised learning process for estimating turbidity. The actual versus estimated curve (Figure 2) also displays the efficiency of the neural network model observing a good fit between predicted and experimental values, as mentioned by Noor et al [30] and Sahin et al [38], which is also explained for the upper value of R 2 , justifying the selection.…”

“…Finally, the neural network was trained with the Levenberg-Marquardt, Bayesian Regularization, and Scaled Conjugated Gradient algorithms. In training each of these algorithms, the number of neurons was modified by increasing five units for each iteration until reaching 100 neurons [29,30]; in total, 57 different models were trained. The backpropagation mechanism, Levenberg-Marquardt, Bayesian Regularization, and Scaled Conjugated Gradient algorithms were used because they are suitable for problems with a large amount of data involving nonlinear relationships.…”