Prediction of Nanofluid Specific Heat Capacity using Supervised Regression Models

Abstract

The accurate prediction of the specific heat capacity (SHC) of nanofluids has become increasingly important for industrial and scientific applications. However, traditional experimental methods of determining SHC are often costly, labor-intensive, and subject to measurement uncertainties. Consequently, there is a need for reliable and efficient predictive models. The aim of this study is to develop and evaluate supervised machine learning regression models capable of predicting the SHC of nanofluids based on their key thermophysical features. A data set of 517 records containing the thermophysical features of nanofluids is collected and preprocessed. The dataset features include nanoparticle type, base fluid, base fluid temperature, and nanoparticle volume fraction. Supervised regression models such as Gradient Boosting, XGBoost, AdaBoost and Decision Tree Regressor, were applied and evaluated. The Gradient Boosting model showed the best performance with R² score of 99.60%, followed by XGBoost (97.43%), AdaBoost (97.07%) and the Decision Tree Regressor (86.73%). These findings demonstrate the capability of machine learning regression models in predicting SHC, offering a cost-effective and rapid alternative to experimental measurements. The results highlight the potential of such approaches to support the design and optimization of advanced thermal management and energy systems.