MOOCs have attracted hundreds of millions of learners with advantages such as being cost-free and having flexible time and space. However, high dropout rates have become the main issue that hinders their further progress. To solve this problem, this research proposes a pipeline model named CLSA to predict the dropout rate based on learners’ behavior data. The CLSA model first uses a convolutional neural network to extract local features and builds feature relations using a kernel strategy. Then, it feeds this high-dimensional vector generated by the CNN to a long short-term memory network to obtain a time-series incorporated vector representation. After that, it employs a static attention mechanism on the vector to obtain an attention weight on each dimension. When tested on the KDD CUP 2015 dataset, our model reached 87.6% accuracy, which was higher than the previous best model (over 2.8%). The F1-score of our model reached 86.9%, which was 1.6% higher than the previous state-of-the-art result.

MOOCs以免费、时间和空间灵活等优势吸引了数亿学习者。然而，高辍学率已成为阻碍其进一步发展的主要问题。为了解决这个问题，本研究提出了一种名为 CLSA 的管道模型，用于根据学习者的行为数据预测辍学率。CLSA 模型首先使用卷积神经网络提取局部特征，并使用内核策略构建特征关系。然后，它将 CNN 生成的这个高维向量提供给一个长短期记忆网络，以获得一个包含时间序列的向量表示。之后，它在向量上采用静态注意力机制来获得每个维度上的注意力权重。在 KDD CUP 2015 数据集上进行测试时，我们的模型达到了 87.6% 的准确率，这高于之前的最佳模型（超过 2.8%）。我们模型的 F1 分数达到了 86.9%，比之前最先进的结果高 1.6%。