An efficient model for the prediction of polymerisation efficiency of nano-composite film using Gaussian processes and Pearson VII universal kernel

Abstract

Polymerisation efficiency of nano-composite film is a very important parameter for film preparation. It is essential to suggest a modelling method to predict and analyse the polymerisation efficiency of nano-composite film. An algorithm combined with Gaussian processes (GP) and Pearson VII universal kernel (PUK) was used in the prediction of polymerisation efficiency of nano-composite film. The input parameters are laser energy density, environmental pressure, laser ablation deposition time, and the distance between target and substrate, while the output parameters is the polymerisation efficiency. In the experiment, the mean absolute error and root mean squared error of GP-PUK model are 14.5142 and 17.2338, respectively, which are smaller than those of GP-poly, GP-normalised poly and GP-RBF models. In order to make further verification to the effectiveness of the model, ten-fold cross validation was used, under the same sample database, to make comparisons between the linear regression (LR), multilayer perceptron (MLP) regressor, radial basis function (RBF), support vector regressor poly kernel (SVR-Ploy) and support vector regressor PUK (SVR-PUK). Comparison results show that the effect of the GP-PUK model in predicting the polymerisation efficiency of nano-composite film is superior to those of the other models.