<p>Thermal conductivity (K) is an important parameter of thermal properties of building materials, in order to improve the high-value resource utilization of waste rubber and optimize the performance of building energy-saving materials, the K of waste rubber-kaolin mixtures (RKM) was tested by indoor thermal detection test, and the effects of rubber particle content, dry density and moisture content on the K of the mixture were studied. The K prediction model is established by artificial neural network technology, and compared with the traditional empirical model. The results show that the K of waste RKM is less affected by rubber content. The K increases with the increase of moisture content, and the critical moisture content is mainly concentrated in 45% ~ 55%. The K of the mixture increases with the increase of dry density. From the neural network prediction model, the correlation coefficient R² of the predicted K of the mixture is greater than 0.91. The theoretical calculation model can accurately calculate the K of the mixture, and its correlation coefficient R² = 0.9178, absolute mean error MAE = 0.084 (W·m⁻¹·K⁻¹).The root mean square error RMSE = 0.106 (W·m⁻¹·K⁻¹).The accuracy of the theoretical calculation prediction model is better than that of the traditional empirical relationship model, which provides an accurate method for predicting the K of waste RKM.</p>