QSPR studies on oxide ion conductivity of perovskites electrolyte based on doped LaGaO_3

Peng Chunrong,Liu Xu,Lu Wencong

DOI: https://doi.org/10.3969/j.issn.1001-4160.2009.11.017

2009-01-01

Abstract:In this paper,Molecular parameters and ionic conductivities data of perovskite-type oxides of doped LaGaO_3 series were collected from literatures and experiments.The relationship between the electrical conductivities and the molecular parameters was examined. The data were processed by several machine learning methods,including Multiple Linear Regression(MLR),Partial Least Squares(PLS) and Support Vector Regression(SVR).The conductivities values of SVR leave-one-out cross validation were corresponded with the values of experiments.The correlation coefficient of calculated values and experimental values was 0.911.Independent test set was used for prediction,and the correlation coefficient of values of prediction and experiment was 0.880.The model was used in the prediction of doped system of La_(1-x)Sr_xGa_(1-y)Mg_yO_3.A diagram of contour was obtained from the results of prediction.It showed excellent agreement with experimental results.