Utilizing Clickstream Data To Reveal The Time Management Of Learners’ Online Self-regulated Learning

Nowadays,online learning has become a widespread and very popular learning mode in higher education which provides learners with self-paced learning without limitation of time and place.With the access of the Internet,online learning breaks the limitations of traditional face-to-face classroom education in terms of time,place and learning rhythm,and enables learners to participate in learning anytime and anywhere.Online learning characterized by autonomy and flexibility has become an important part of higher education.Although online learners can complete course tasks according to their favorite rhythm and learning style,they need to independently plan,monitor and reflect on their learning process,that is,they need to carry out self-regulated learning(SRL).Existing studies have shown that the differences in learners’ academic performance are not caused by intelligence or diligence,but mainly due to the differences in SRL strategies.Time management strategy is a very important component of SRL framework,which is closely related to organizing learning time to achieve the expected academic goals.Based on the clickstream data of 8019 learners recorded in the starC system log in the spring semester of 2019,this study is committed to utilizing the clickstream data to reveal the time management of learners’ online self-regulated learning in a large-scale real online learning environment.This study explores the specific differences in time management(time investment and time use patterns)of self-regulated learning among learners with different academic performance categories(high,medium and low)and different departments(science,literature,arts and sports),and uses the convolution neural network CNN model to recognize learners with different academic performance categories in the middle of the term based on the "time management images".The specific research contents and contributions of this study are summarized as follows:Firstly,the study on the difference of learners’ time investment of SRL in online learning.Time investment is one of the key attributes of time management strategy.Time investment mainly refers to the time spent by learners in SRL in the real online learning environment of higher education.In this study,the total number of days and duration of SRL in online learning in a term are considered as typical indicators to measure time investment.Based on whether learners participate in online SRL every week,this study divides how learners regulate their time investment in a semester into four types of time investment(continuous multi investment,more first and then less investment,continuous less investment,less first and then more investment).The research results show that there are differences in the total number of days and total duration of online SRL in a semester among learners with different academic performance categories and departments,and there are also significant differences in the proportion of the four types of time investment among learners with different categories.The results of this study confirm that it is difficult to achieve ideal academic performance without sufficient time investment,the adequate time investment is very important for the success of flexible and autonomous online learning.Secondly,the study on the difference of learners’ time use patterns of SRL in online learning.Time use patterns are another key attribute of time management strategy,which mainly refers to the time selection and allocation of learners to complete learning tasks.This study considers not only the time choice of learners’ SRL,but also the time sequence of their SRL in online learning,that is,the study regularity related to time management strategy.In this study,time use patterns were measured by the percentage of students participating in SRL every day in the term in online learning,the percentage of records participating in SRL every hour in a day in online learning,and the study regularity of participating in SRL in the term in online learning calculated by true entropy.The analysis results of this study will help to better reveal the time choice and sequence of SRL for learners with different academic performance categories and departments throughout the semester and a day in the online learning environment of higher education.We will have an in-depth understanding of the time use patterns of learners with different categories.Thirdly,the study of learner recognition based on "time management images".The rapid development of deep learning(DL)and its application in education,especially the outstanding performance of convolutional neural network(CNN)model in image recognition,provides an opportunity to effectively recognize learners with different academic performance categories in the early stage of the learning process.SRL before the middle of the term(the ninth week)was converted into single channel and three channel "time management images",and the CNN model of learner recognition was built with reference to the digit recognition model.This study aims to explore whether the CNN model based on "time management picture" can effectively recognize learners with different academic performance categories in the middle of the term.The results showed that the accuracy of recognizing learners based on the three channel "time management images" is slightly higher than that based on the single channel "time management images",and the recall rate of recognizing learners based on the three channel "time management images" is significantly higher than that based on the single channel "time management images",while the recognition effect for science learners is the best.In this study,the clickstream data of learners’ SRL was converted into"time management images",and the CNN model is used to recognize learners in the middle of the term based on "time management images",which provides a new way for the recognition of learners.In the large-scale real online learning environment of higher education,this study relies on the clickstream data collected in the starC system log of H University in a way that does not affect the normal teaching order to explore the differences in time management(time investment and time use patterns)of learners with different academic performance categories and different departments,and uses the CNN model to recognize learners in the middle of the term based on the "time management images".This study verifies the differences of time management strategies among learners with different academic performance categories and departments,which can deepen the theoretical research of SRL.The results of this study provide a new perspective for education managers and teachers to consider the factors that lead to the differences in learners’ academic performance.They can attribute the differences in academic performance of learners to the controllable factor of time management,and then focus on providing targeted interventions to optimize learning process and improve overall performance.The main deficiency of this research is that click stream data is only the external behavior of learners in the learning process,without considering the internal subjective perception and other modal data of learners in the learning process.Future research work should combine specific teaching situations and curriculum design,and try to integrate multimodal data to carry out research on the automatic identification,understanding and support of learners’ SRL process,so as to create a personalized learning environment for learners,promote the effective implementation of learners’ learning,and finally truly realize intelligent and personalized learning.