Student score prediction method based on depth matrix decomposition

Abstract

The invention provides a student score prediction method based on depth matrix decomposition. The method is characterized in that the test score of each chapter, the class assignment score of each chapter and the final score of the course of a certain professional course of the subject are used as input data; and input data is decomposed into student features and project features through a matrixdecomposition method, then learning is carried out through a forward propagation full-connection neural network, and simple features are combined into more complex features. According to the method, the full-connection neural network with an attention mechanism is designed, the student potential feature vectors and the project potential feature vectors are obtained by constructing the feature vectors of the students and the projects, and finally the cosine similarity is calculated to obtain the prediction score, so that the accuracy and the interpretability of the prediction result are improved.

Description

[0001] (1) Technical field [0002] The invention relates to technical fields such as machine learning, recommendation algorithms, and data mining, and in particular to a method for predicting student performance based on deep matrix decomposition. [0003] (2) Background technology [0004] In the era of big data, massive educational data has become one of the key resources for schools, teachers and other fields related to education and learning. People need to discover knowledge information hidden in educational data through data mining. Subsequently, educational data mining has become an emerging interdisciplinary field, which uses data mining techniques to analyze and extract hidden knowledge from educational data to solve problems in teaching practice. In educational data mining, student performance prediction is an important issue in education management. By predicting student performance, it can help students with learning difficulties in a timely manner, and at the sam...

AI Technical Summary

Problems solved by technology

[0005] In order to realize the prediction of students' grades, the invention patent "A Method and System for Predicting Students' Academic Grades Based on BP Neural Network Model", the publication number is CN106157204A, mainly uses the academic grades of students approaching two semesters and the data of entrance academic grades. There are technical problems in the prediction of students' academic performance, and the intelligent prediction of students' academic performance is realized, but the feature granularity extracted is relatively coarse, and the prediction results lack interpretability

Invention patent "A Method and System for Predicting Students' Academic Performance Based on Naive Bayesian Model", the publication number is CN106127634A, which mainly converts the data of students' learning data stored in the database to obtain a standardized data table of students' learning status ;Aiming at the standardized student learning status data table, calculate the conditional probability of each attribute in different categories and the prior probability of different categories through the computing unit in the server; convert the student data to be predicted and input it into the trained simple shell The Yaesian model predicts the classification of student data, but the Naive Bayesian algorithm needs to know the prior probability, and the prior probability often depends on the assumption, and the algorithm is very sensitive to the expression form of the input data

Images