Multi-label prediction method and system for comprehensive scores of students

Abstract

The invention discloses a multi-label prediction method and system for comprehensive scores of students, and the method comprises the steps that an attention weight value between each group of past test scores and each group of learning behavior label data is calculated through constructing a plurality of groups of attention mechanisms; according to the attention weight value, feature vectors between each group of past test scores and the multiple groups of learning behavior label data are calculated until multiple groups of feature vectors are obtained, and each group of feature vectors fully reflect the incidence relation between the multiple groups of learning behavior labels and one group of past test scores; complementation of key information between past test scores and learning behavior label data is realized; the fused feature vector fully considers the degree of action of different learning behavior labels on the next test score of the student and the degree of action of different previous test scores on the next test score of the student, and finally the fused feature vector is input into the trained BP neural network. The accuracy and rationality of predicting the examination scores of the students by the BP neural network are improved.

Description

technical field [0001] The invention relates to the technical field of student grade prediction, in particular to a multi-label prediction method and system for students' comprehensive grades. Background technique [0002] At present, among the various teaching tasks in most colleges and universities in our country, the core of their work is carried out around the management of teaching. In the teaching management work, the core content is the management of student performance. [0003] In related schemes, a neural network model has been constructed, and then past test scores and learning behavior labels that affect test scores in the semester are input into the trained neural network model to obtain the prediction of the model output for the next test of the students Although the related scheme can predict the students' next test scores, the scheme is to input irregular discrete data into the neural network model, ignoring the different effects of different learning behavi...

AI Technical Summary

Problems solved by technology

[0003] In related schemes, a neural network model has been constructed, and then past test scores and learning behavior labels that affect test scores in the semester are input into the trained neural network model to obtain the prediction of the model output for the next test of the students Although the related scheme can predict the students' next test scores, the scheme is to input irregular discrete data into the neural network model, ignoring the different effects of different learning behavior labels on the next student's test scores and the different effects of different past test scores on students' next test scores, and also ignores the correlation between different past test scores and different learning behavior labels, making the prediction accuracy low

Images