Classification of nanofluids solutions based on viscosity values: A comparative study of random forest, logistic model tree, Bayesian network, and support vector machine models

Abstract

In this paper, we investigated predictive modelling algorithms namely Logistic Model Tree (LMT), Bayesian Network (BN), Random Forest (RF), and Support Vector Machine (SVM) for classification of nanofluid solutions based on viscosity values in the presence of different concentration of salinity, silica nanoparticles, and polyacrylamide solutions. The performance of these classification models was verified through validation and testing data sets. As an initial evaluation, these models were used to classify nanofluid solutions into four classes based on viscosity values. The results obtained in this study can also be useful in establishing comparative analysis with different types of preprocessing methods such as centering, normalization, standardized and firstorder. After comparing the results of the four models of classification with each of preprocessing methods, the result of the nanofluid classification was acceptable. According to the classification results, the firstorder-BN model provided better results than the normalized RF model, and the best result of classification was obtained by LMT for nanofluid solutions. Using LMT, the Mean Absolute Error (MAE) and Root Mean Square Error (RMSE) for all tests with different preprocessing methods were 0.0709 and 0.2594, respectively. The proposed method is rapid to measure the viscosity values of nanofluid solutions in oil industry for polymer flooding in chemical enhanced oil recovery (C-EOR).