Integrated Algorithm of Stacking and its Application in Prediction

Ensemble learning constructs and combines multiple learners to complete learning tasks. This paper studies the XGBoost model, logistic regression model, and LightGBM model based on the integrated Stacking algorithm model. XGBoost model can further realize the cross combination of features through serial results, and obtain the global optimal parameter combination within the grid parameters. Logistic regression model to analyze the predicted value of all types of data, including continuous variables, discrete variables and dummy variables, without considering the prior probability loss; the LightGBM model filters the original features using the LightGBM feature selection algorithm, and then truncates the feature variables according to different threshold conditions. In order to verify the applicability of the integrated Stacking algorithm model constructed in this paper, student learning behavior features were constructed, and Pearson coefficient was used to calculate the correlation between the features. The experimental results show that the prediction model constructed with the integrated Stacking algorithm can meet the requirements of predicting scores in universities.