Abstract: There are a large number of online learning courses and there is a clear problem of information overload. One of the most effective ways to solve the problem of information overload is to use personalized recommendation system. According to the characteristics that the learners’ courses are often time-sequential, an online course recommendation model based on LSTM (long short-term memory) network is proposed. The characteristics of learning behaviors are extracted from the sequence of lessons learned by a large number of learners, thereby predicting the courses that the learner will learn. The algorithm proposed in this paper is based on the time sequence between courses. Therefore, according to the closeness of the relationship between courses, the accuracy of the recommendation after course classification is higher. Due to the continuous update of online courses, the workload of manually maintaining the course classification is large, and the classification is not scientific enough. This paper uses GSP (generalized sequential pattern mining algorithm) and spectral clustering algorithm to discover the hidden time linkage between courses and proposes a more reasonable course classification method. Compared with the traditional collaborative filtering algorithm and the course recommendation algorithm based on RNN (recurrent neural network), experimental results show that the accuracy of the proposed algorithm is higher.

Key words: intelligent recommendation, course sequence, deep learning, long short-term memory (LSTM) network, data mining

摘要： 在线学习课程数量庞大，存在明显的信息过载问题，个性化智能推荐是解决这一问题的有效方式。根据学习者所学习的课程往往具有时间序列性这一特点，提出了基于LSTM网络的在线课程推荐模型。从大量学习者所学习的课程序列中提取学习行为特点，进而预测学习者将要学习的课程。该算法是基于课程之间的时序性而提出的，因此按照课程之间关系的紧密程度将课程分类后推荐的准确率更高。由于在线课程不断更新，人工维护课程分类的工作量较大，同时分类也不够科学，利用GSP算法和谱聚类算法，挖掘出课程间隐藏的时序联系，提出了更合理的课程自动分类方法。实验结果证明，与传统的协同过滤算法以及基于RNN的课程推荐算法相比，该算法推荐准确度更高。

关键词: 智能推荐, 课程序列, 深度学习, 长短时记忆（LSTM）网络, 数据挖掘