ONLINE BEHAVIOR PREDICTION BASED ON DEEP LEARNING IN HEALTHCARE

Abstract

In recent years, with the rapid development of Internet and computer technology, network education has developed rapidly. With the rapid development of learning technology, online education has been widely popularized. Especially in 2020, novel coronavirus pneumonia suddenly came into being. Online education based on Internet technology has played a great role in the crisis control period. It has also enriched teaching forms and teaching methods. The blended teaching under online and offline integration has increased the availability of students’ learning data. Therefore, more and more scholars begin to pay attention to the research of learning early warning based on educational data mining or learning analysis. However, most early warning studies use traditional machine learning algorithms, and there are still deficiencies in the granularity of data collection, technical implementation mechanism, early warning state recognition and so on. With the success of deep learning in artificial intelligence and other fields, scholars began to study the application of deep learning to solve the problems in the field of learning early warning. Combining variational self-coding (LVAE) and deep neural network, this paper proposes a scheme (LVAEpre) which can solve the problem of unbalanced distribution of educational data sets. This paper determines the weight value of each dimension and index by adjusting the weight parameters of the model, and obtains the threshold value of the early warning line, and empirically tests its effectiveness. Finally, the paper designs a learning early warning model and builds a learning early warning platform based on process data. The results show that the early warning effect is good. The proposal of the learning early warning model based on process data and the application of the learning early warning platform have greatly improved the teaching quality, reduced the risk of students failing to attend the course, and effectively realized the early warning function. The experimental results show that the framework improves the prediction ability of identifying risk learners as soon as possible, timely intervene and guide risk learners, improves learning efficiency, and provides effective guidance strategies for the development of network education.