<p>The burn through point temperature is a key parameter that affects sintering production. In order to predict the endpoint temperature and take adjustment measures, this paper conducted a study on interpretable dynamic prediction and optimization of burn through point temperature. Data from the sintering process were collected. Anomalies and time delays in the data were treated. Feature selection was performed using Extremely Randomized Trees and mean squared error loss function. The Extreme Randomized Trees burn through point temperature prediction model was established. <i>R</i>² was 0.85, and the predicted hit rate was 88%. Interpretable analysis was performed on burn through point temperature prediction model using different methods such as partial dependence plot, individual conditional expectation, accumulated local effects, local interpretable model-agnostic explanations, shapley additive explanations, etc. The trend effect of the operational parameters was obtained. The appropriate range of operational parameters under different conditions was studied. The prediction process of a single sample was analyzed. State parameters prediction models were established and single sample analysis was conducted. Obtained operational parameters that have a significant effect on the state parameters. By adjusting operational parameters and optimizing state parameters, the burn through point temperature was increased from 415.91&#xa0;℃ to 465.82&#xa0;℃.</p> Graphical Abstract <p></p>