A student dropout prediction method based on attention deep learning model

Abstract

The invention discloses a method for predicting student dropouts based on an Attention deep learning model, which includes encoding and time windowing of behavior records to generate multiple time slice behavior vector matrices; the improved convolution layer based on CNN network is multi- A time slice behavior vector matrix is ​​processed to obtain multiple eigenvector matrices with local correlations; based on the BI‑GRU model, multiple feature vector matrices with local correlations are extracted for time-series characteristics, and multiple time-series relationships are generated. Behavior feature vector matrix; based on the Attention mechanism, assign different weights to the hidden layer features at each time, weighted and sum the hidden layer states and weights at different times, generate behavior feature representation vectors, and input them into the classification layer, and use the Sigmoid function to perform Forecast and get the forecast result. By considering the relationship between students' learning behaviors and the influence of different behavioral characteristics on dropout prediction, the prediction is realized and the accuracy of dropout prediction is improved.

Description

technical field [0001] The invention relates to the technical fields of machine learning, deep learning and data mining, in particular to a student dropout prediction method based on an Attention deep learning model. Background technique [0002] With the rapid development of the Internet, the educational model has begun to change, and more and more different groups of people have different demands for educational quality and educational content. Therefore, a large number of online course learning platforms have emerged as the times require. Since MOOC, a large online learning platform, has emerged since 2012, it includes famous schools from all over the world publishing courses on MOOC. The learning platform transcends the limitations of time, space and even identity. As long as people who are willing to learn, no matter what profession they are engaged in, they can conduct online learning by registering an account, but this also brings serious problems. The dropout rate is...

AI Technical Summary

Problems solved by technology

[0003] The purpose of the present invention is to provide a student dropout prediction method based on the Attention deep learning model, aiming to solve the problem of low accuracy of the traditional dropout prediction method

Images