Student Achievement Prediction Based on NN-Stacking Model with Hyperparameter Optimization

In terms of student achievement prediction, this paper proposes the use of random search algorithm to optimize the NN-Stacking model with hyperparameters. Compared with using grid search algorithm to configure hyperparameters, using random search algorithm can more efficiently select suitable hyperparameters for the model. Compared with Ridge regression, SVR, Random Forest, and Adaboost data mining algorithms, the optimized RG-UNNS and RG-CNNS fusion algorithms significantly improve the accuracy of student achievement prediction, and the MSE decreases 18.2% and 10.57%, respectively, compared with the linear Stacking algorithm. RG-UNNS has the best prediction effect when the time requirement is not high.