Date of Publication :11th May 2021

Abstract: MOOC is an ultimate way to give educational content in higher studies settings by providing good-quality educational material to the studentts throughout the world. If we take in the differences between traditional learning paradigm and MOOCs, a trend focusing on predicting and explaining dropping out of students and low completion rates in MOOCs has emerged. The quantity of students taking these courses is very high, still, the completion rate is very low. Factors affecting the completion of a course by a student such as interest in the subject, reason for taking the course, if the teacher is able properly make students understand or not. Nevertheless, because of varying problems specifications and evaluation metrics, undertaking a high level evaluation of state-of-the-art machine learning architectures isn't easy. This paper provides a complete rundownn of MOOCs student's dropout probability with the help of machine learning techniques. Moreover, we have highlyghted a few answers being used to solve the dropout problem, provide an examinationof the challenge of prediction models, and give some important overview and suggestions that may pave the way to develop useful ML solutions for overcoming the MOOCs dropout problem. Essentially, we got features from every learning behaviors and formed multi-view behavior features. Also, examining these features, we proposed a new multi-view semi-supervised learning model to use a large number of raw data to help inadequate labeled data to improve prediction. We have used KDD Cup 2015 dataset, from the results we see that our proposed model attains better prediction of a student leaving a course than state-of-the-art methods. Long Short-Term Memory neural network (LSTM) prediction model makes use of time-control units, the unit has the ability to model early learning behaveiour with varying time intervals. Formulated from the LSTM model, we designed time-controlled gates for capturing a good long- and short term info and use the learning process info toget better forecast performance

Reference :