Achievement Prediction System Based on Stacking Integration Algorithm

The Stacking algorithm is a heterogeneous ensemble learning method that combines learning methods as a combination strategy. The first layer classifier is called the base classifier, and the second layer classifier is called the secondary classifier or meta classifier. The score prediction model based on Stacking ensemble algorithm consists of three parts: the first part preprocesses the data, the second part trains the base regressor, and the third part outputs the prediction results. The base regressor consists of LightGBM, XGBoost, AdaBoost, CatBoost, etc. The four base regressors use the same feature samples for training, and the training process of each base regressor does not affect each other. Finally, 4393 student data were collected for simulation experiments to verify the superiority of the performance prediction model based on Stacking ensemble algorithm. The advantage of Stacking ensemble learning algorithm is to combine multiple algorithm models, reduce overfitting problems of a single model, and improve the generalization ability of the model.