It is important to determine the exact value of the overburden load on the tunnel roof to assess the stability of the roof, ho wever, th is parameter is poorly constrained. Based on this, we use the hybrid intelligence approach to deal with the nonlinear relationship between multiple parameters and the overlying load. Firstly, a multi parameter correlation visualization analysis is performed to determine two hidden layers (E 1 , E 2 ), and four parameters with the most influence on the overlying load were selected, which will be used as sample input parameters. Secondly, the top plate stability prediction is achieved by converting the input and output to a single thermal code, and then adding parameter thresholds that were continuously adjusted with training . This completes the structure of ANN IHHO model . Finally, the model is trained based on data driven by displacement inverse analysis and numerical experiments. The model prediction accuracy was then validated using a confusion matrix form . The validation results show that the trained ANN IHHO model has good accuracy and can predict the top plate stability under the condition of considering four parameters The proposed model provides a new way of thinking to determine the overburden load and stability of the top plate . This study proposes many new research ideas in the process of developing the model : First, the m ulti parameter correlation visualization analysis can be used in studies to assess the degree of correlation between parameters and results, also between parameters and parameters And most importantly, the visualization analysis can intuitively display the correlation degree . the introduction of displacement inverse analysis as a data acquisition method into the ANN model has proven to be feasible, and the method has the advantage of being low cost and easy to implement ; Third, the c onfusion matrix also shows good adaptability in verifying model prediction accuracy, especially for prediction models of rock stability The method proposed in this study to develop the model is also adapted to the prediction of other rock mechanical parameters.