The article is devoted to the use of neural networks for predicting the mechanical properties of rubber. Rubbers include, as a rule, more than one and a half dozen components. Each of the components has a complex and ambiguous effect on the complex of material properties. When developing new compositions, this significantly complicates and lengthens the solution of material science problems by traditional methods of composition selection. These problems can be effectively solved using machine learning techniques. The authors have developed approaches to the use of neural networks for predicting the mechanical properties of rubber from a known composition. In this article, neural network models have been created and optimized, which make it possible to predict the mechanical properties of elastomeric materials with high accuracy.