Loading [MathJax]/extensions/TeX/ieeemacros.js
Multi-model Stacking Ensemble Learning for Student Achievement Prediction | IEEE Conference Publication | IEEE Xplore

Multi-model Stacking Ensemble Learning for Student Achievement Prediction

Publisher: IEEE

Abstract:

Educational data mining has always been a direction of big data applications, among which the student achievement prediction is one of the research focuses. The research ...View more

Abstract:

Educational data mining has always been a direction of big data applications, among which the student achievement prediction is one of the research focuses. The research has two main objects, one is online education, and the other is school education. Due to the prevalence of online education and the relatively easy collection of data, many studies have focused on the prediction of student achievement in online education. At the same time, the differences in teaching methods, student habits, and technical development levels of schools in different regions have led to a slight deviation in the main research areas and research emphasis on student achievement prediction. In China, due to the difficulties of traditional schools in collecting student academic data, relevant research can only use limited grade data and campus card data. This method has limitations that are not easy to promote. This paper uses a more easily collected score data, combined with multi-model stacking ensemble learning (MMSE) model to predict students' achievement. The model mainly includes two parts: data preprocessing and model construction. In the data preprocessing part, the data features are designed using the students' grades and course information. In the model building part, a two-layer integrated learning model is established. In the first layer, 5-fold cross-validation is used to train 5 different basic classifiers. The second layer uses the regression algorithm combined with the prediction results of the first layer to predict student achievement. The results show that using the data combined with appropriate feature engineering to make predictions can get very satisfactory results.
Date of Conference: 10-12 December 2021
Date Added to IEEE Xplore: 04 March 2022
ISBN Information:
Publisher: IEEE
Conference Location: Xi'an, China

References

References is not available for this document.