语义增强的在线学习行为预测研究

当前学习者的在线学习行为预测研究未充分利用短文本中的语义数据,导致对学习者的学习状态刻画不够全面,严重影响了行为预测的准确性.针对此问题,本文提出了语义增强的在线学习行为预测方法.首先,利用双向长短时记忆网络得到到短文本的语义向量表示;其次,基于学习者的统计、行为和短文本数据得到学习者的特征表征,并利用长短时记忆网络模型构建其学习状态表征;最后,利用学习状态表征预测学习者的学习行为.在11门真实在线课程数据集上的实验表明,本文方法能过有效提升在线学习行为预测的精确度.

Semantic Enhanced Behavior Prediction Method for Online Learners

The current learning behavior prediction research for online learner does not make full use of the semantic data in the short text,which leads to the incompleteness of the learner’s learning state,which seriously affects the accuracy of behavior prediction.To solve this problem,this paper proposes a semantic enhanced online learning behavior prediction method.Firstly,the BiLSTMmodel is used to obtain the semantic vector representation of short text.Secondly,the learner’s feature representation is obtained based on the learner’s statistics,behavior and short text data,and the learning state representation is constructed by using LSTMmodel.Finally,the learning state is used to represent the prediction learning behavior.Experiments on 11 real online course datasets showthat this method can effectively improve the accuracy of online learning behavior prediction.