基于改进LSTM网络的自主学习行为分析预测  

Analysis and prediction of autonomous learning behavior based on improved LSTM network

作　　者：郎波 樊一娜 姚建欣 李春密[3] LANG Bo;FAN Yi-na;YAO Jian-xin;LI Chun-mi(Zhuhai City Polytechnic,Zhuhai 519090,Guangdong Province,China;Zhuhai Campus of Beijing Normal University,Zhuhai 519087,Guangdong Province,China;Beijing Normal University,Beijing 100875,China)

机构地区：[1]珠海城市职业技术学院,广东珠海519090 [2]北京师范大学珠海校区,广东珠海519087 [3]北京师范大学,北京100875

出　　处：《信息技术》2024年第11期84-91,共8页Information Technology

基　　金：中国科普研究所重点招标课题(220101ESR002);国家自然科学基金应急管理项目(71641016);珠海市教育科研“十四五”规划课题(2021ZHGHKTG183);珠海城市职业技术学院课题(ZLJS20220509,KCSZ-20200605)。

摘　　要：为了实现在有限时间内获取学习效果的最优化,文中提出了一种改进的基于长短时记忆神经网络的学习行为分析预测模型。首先利用经验模态分解对学习数据进行特征提取,从频率上将其分解为多个本征模态函数,然后利用K均值聚类方法对分解后的分量进行集成,从中选取最优聚类数生成训练数据,将其输入到长短时记忆神经网络中进行特征融合,从而实现对学习行为的分析预测。从实验结果分析,该方法与其他模型对比具有较高的预测精度,能够较准确地实现对学生学习行为的预测。In order to optimize the learning effect in a limited time,this paper proposes an improved learning behavior analysis and prediction model based on Long Short Term Memory(LSTM)neural network.Firstly,empirical mode decomposition is used to extract features from the learning data,decompose it into multiple intrinsic mode functions from the frequency.Then,the K-means clustering method is used to integrate the decomposed components,among which the optimal clusters number to generate training data.Nextly,the training data would be input into the LSTM memory neural network,so as to realize the analysis and prediction of learning behavior.From the analysis of experiment results,this method has higher prediction accuracy compared with other models,and can accurately realize the analysis and prediction of students’learning behavior.

关 键 词：自主学习行为 经验模态分解 聚类 长短时记忆神经网络 分析预测 

分 类 号：TP181[自动化与计算机技术—控制理论与控制工程] TN912.3[自动化与计算机技术—控制科学与工程]