中国电化教育

大数据云计算平台触发了教育行业的巨大变革,催生了各种形式和各种内容的网络教学开放课程,这些课程所采用的自由在线学习方式能够针对学习者的个性化需求展开导向学习,可以激发学习者的学习积极性。而如何根据学习者的特点进行较为准确的评价是网络教学开放课堂所面临的一个关键问题。该文利用在线学习过程中学习者学习活动所产生的大量数据对学习者产生形成性评价,首先基于认知思维的要求作为在线学习的目标分类,通过自动分类算法进行聚类分析区分学习者的应用能力层次。在此基础上,为提高数据的处理效率,加快对学习者的评价,再采用三层自动编码器的神经网络降维提取关键特征,然后利用训练集中的数据训练学习器,并利用集成学习机制综合单个学习器的结果,得到更为准确的评价结果。该文采集初中学生针对特定知识点的在线学习活动数据构成训练数据和测试数据集,通过对比专家人工和机器学习机制对所构造的测试集数据的评价结果,验证了该文方法的有效性。

Cloud computing platform supported by big data has triggered the significant revolution of education, generating all kinds of open courses of online teaching. These open courses can guide the educatees' study and meet the educatees' individual requirements by using free online learning mode, stimulating the educatees' study enthusiasm. For open courses of online learning, it is a critical problem to give the accurate evaluation according the educatees' individual characteristics. To address the issue, this paper proposes a formative assessment by utilizing the numerous online learning data of educatees' learning activities on open courses. Firstly, taking the cognitive hierarchy as the classification standard of online learning, the educatees' knowledge applicable level is classified by k-means clustering analysis. Then to improve the efficiency of data processing and accelerate the evaluation, the data feature is extracted by using three-layer automatic encoder based on neural network to reduce the data dimension and obtain the key features. After feature extraction, the dimension-reduced data in training set is utilized to train the individual machine learner. Then the integrated learning mechanism is introduced to synthesize the learning result of every individual machine learner, making the evaluation results more accurate. In this paper, the data of online learning activity of junior middle school students on specific knowledge points is collected to construct the training data and test data sets. The effectiveness of the proposed method is verifies by comparing the formative assessment results with the experts' manual evaluation on the test data set.