Exploring the impact of different clustering techniques and their parameters on the precision of viscosity predictions for water-based nanofluids containing magnetic nanoparticles

Artificial neural networks (ANNs) are widely used for modeling and predicting complex experimental systems due to their strong ability to capture nonlinear relationships. Among them, radial basis function (RBF) networks are particularly effective; however, their performance strongly depends on the proper selection of centers before training. To address this limitation, clustering techniques can be used to determine optimal centers. In the context of nanofluid viscosity prediction, integrating clustering methods with RBF networks enhances model accuracy and represents an effective approach for improving predictive performance. This research investigates the influence of various clustering methods, including Self-Organizing Map (SOM), K-means, K-medoids, Fuzzy C-means (FCM), and MATLAB's clustering function, on the precision of viscosity prediction. By integrating clustering algorithms and MATLAB's cluster function with RBF networks, researchers can leverage the complementary strengths of both approaches to enhance the overall performance of curve-fitting and prediction models. Among the investigated methods, the FCM-based model achieved the highest accuracy, with an MSE of approximately 1.2 × 10 ^-3 and maximum and minimum error percentages below 4% and 1.5%, respectively. The collaboration between clustering methods and RBF networks fosters a more intricate understanding of the complex relationships within data sets, thereby enabling more precise and dependable viscosity estimations for nanofluid systems. The outcomes of this analysis can provide valuable insights into the optimal clustering method to use alongside viscosity prediction models, potentially enhancing efficiency.