نوع مقاله : ترویجی
عنوان مقاله English
نویسندگان English
This study focuses on predicting the specific heat of polymers using experimental data and machine learning algorithms. A comprehensive dataset comprising physical and chemical properties of polymers was utilized as input for various models, including linear regression, Ridge, Lasso, ElasticNet, K-nearest neighbors, artificial neural networks, and decision trees. The performance of these models was evaluated using mean squared error (MSE) and R-squared (R²) metrics. The results indicate that the Lasso and ElasticNet models provide the best balance between accuracy and generalization to new data. Specifically, the Lasso model achieved an MSE of 728.77 and an R² of 0.95, while the ElasticNet model recorded an MSE of 604.34 and an R² of 0.95. This study emphasizes that employing regularization techniques and ensemble models can significantly enhance the accuracy and stability of predictions for the thermal properties of polymers, leading to substantial savings in both time and costs associated with experimental testing.
کلیدواژهها English
