基于双注意力机制的成绩预测  

Performance Prediction based on Dual-Attention Mechanism

作　　者：张文娟 张彬[1] 杨皓哲 Zhang Wenjuan;Zhang Bin;Yang Haozhe(School of Mechanical Engineering,Tongji University,Shanghai 201800,China)

机构地区：[1]同济大学机械与能源工程学院,上海201800

出　　处：《南京师大学报（自然科学版）》2023年第4期103-113,共11页Journal of Nanjing Normal University(Natural Science Edition)

基　　金：大型交通枢纽智能协同运营关键技术研究与示范(21DZ1203700).

摘　　要：为更好利用虚拟学习环境中的数据,提升学习者成绩和教师教学效果,准确预测学习者的成绩、确定重要影响因素以及确定合适的反馈时间点这三项工作非常重要.本文基于问题域的形式化描述,在明确研究对象特征和假设条件后,提出了一种集成了双注意力机制、门控循环单元(gated recurrent unit,GRU)与一维卷积神经网络(convolutional neural networks,CNN)的网络模型,并在两个公开数据集上进行实验验证.结果显示该模型可以有效实现上述三个核心功能,且在寻找合适的反馈时间时比目前主流方法更为快捷,结果更具普遍性.In order to make better use of the data in the virtual learning environment and improve the performance of learners and the teaching effect,a neural network integrating the dual-attention mechanism GRU(Gated Recurrent Unit)and one-dimensional CNN(Convolutional Neural Networks)was proposed.First,formalized description of the problem domain is given,and some assumptions about the research object are determined.Then,according to the three core functions of the problem,learners'performance prediction,the determination of important influencing factors and the determination of feedback time,a neural network model is constructed.Finally,the experimental results on two public data sets show that the proposed model can effectively realize the three core functions.In addition,it is faster than the current mainstream methods in determining the feedback time,and the results are more general.

关 键 词：计算机应用 虚拟学习环境 成绩预测 反馈时间 注意力机制 神经网络 

分 类 号：TP391[自动化与计算机技术—计算机应用技术]