Photocatalytic Activity Prediction of Rare Earth Doped TiO2 Based on Pattern Recognition

In present paper, the linear regression, Gaussian process regression and support vector machines regression, k-nearest neighbor algorithm were used to predict the photocatalytic reaction rate constant of rare earth doped TiO2. Then, experiments were used to verify the key factor analysis effectiveness based on stepwise multiple regression and advantages of the key factor analysis based on k-NN regression model. Results showed that the prediction performance is better with key factors got from stepwise regression analysis. Compared with other regression methods, it may be found the prediction performance of k-NN is the best. However, the model of support vector regression and Gaussian process regression is much more complex which needs a large number of data as calculation foundation, so the calculation performance was a little lower with fewer basic data points in this article.