Performance evaluation and comparison of machine learning algorithms for prediction of electrodeposited copper ions

Abstract

ABSTRACT Machine-learning prediction algorithms have been used in different scenarios by earlier researchers. This article found the application of these algorithms for predictions of the material deposition rate of copper ions through high-speed selective jet electrodeposition (HSSJED). The presented article emphasises the selection of the best suitable machine-learning prediction algorithms among Gaussian Process Regression (GPR), Support Vector Machine (SVM) regression and Linear Regression (LR) for prediction and compares them with each other for selection of the best algorithm which agrees with experimental data. The comparison of performances of these machine-learning algorithms has been done taking root-mean-square error (RMSE), R2, mean-squared error (MSE) and mean absolute error (MAE) as a basic measure for the same, also various graphs have been plotted and discussed for better understanding. The data set was collected by performing the HSSJED experiment on various parameters like electrolyte composition, electrode gap and applied dc potential. K-fold cross-validation method has been adopted for validation. GPR predictions were found better than the other two which also agree with the experimental output as the quantitative comparison has been done and results are discussed. The average values of RMSE, R2, MSE and MAE were 0.057668, 0.96, 0.0032432 and 0.045733 for GPR.