Skip to main content

A novel nonlinear regression model of SVR as a substitute for ANN to predict conductivity of MWCNT-CuO/water hybrid nanofluid based on empirical data


Arash Karimipour, Seyed Amin Bagherzadeh, Abdolmajid Taghipour, Ali Abdollahi, Mohammad Reza Safaei*

Source title: 
Physica A: Statistical Mechanics and its Applications, 521: 89-97, 2019 (ISI)
Academic year of acceptance: 

An ideal regression method should have several characteristics including precision, accuracy and generalization. In many studies in the field of nanofluid, the precision of models is more highlighted. Nevertheless, a lack of generalization may lead to over fitted models. In this paper, two nonlinear regression methods, namely the ANN and SVR are employed to predict the thermal conductivity of MWCNT-CuO/water hybrid nanofluid with temperature and volume fraction. It is seen that precision of SVR & ANN approaches are able to be compared. However, SVR generalization is more convenient, compared to ANN because SVR method utilizes less parameters. Hence SVR can show better persistence to overfitting in little-size datasets compared to ANN. Therefore, SVR is more authentic approach for the regression with little-size datasets.