Exploration Of The Evaluation Method Of Online Learning Engagement Based On Multisource Data Fusion

As a result of technological advancement,online learning has developed rapidly in the past decade and become a very important part of the current multi-level education system.Although online learning has many advantages,such as maximizing resource utilization,independence of learning behavior,diversity of interaction forms and liberalization of teaching time and space.It also encounters problems of lower learning effect,low retention rate and high dropout rate,which are more common than traditional education.It is necessary for students to have a high level of learning engagement to achieve ideal learning effect.In traditional learning context,teachers evaluate learners’ learning input and intervene or adjust their learning behavior in time to ensure learning effect.Because traditional learning and online learning are essentially consistent,many researchers and educators also hope to solve the problems of low learning performance and high dropout rate by evaluating students’ online learning engagement in online learning context.However,due to the influence of technology and student size,effective methods of measuring and evaluating learning engagement in traditional learning context,such as observation and questionnaire,are also faced with many limitations when applied to online learning situations,which make it difficult to achieve better results.To solve this problem,it is an effective solution to evaluate students’ online learning engagement by using objective data generated in the online learning process.Current researches mostly build evaluation model based on data generated by logging system.Considering the limitation of log data in representing multi-dimensional information of online learning engagement,this study proposed a multi-source data evaluation model that integrates log data,camera data and mouse movement data.The main works of this study include the following four aspects:Firstly,an online learning engagement evaluation model was proposed.Because the three-dimensional model of behavior,emotion and cognition in traditional learning context payed less attention to social dimension.This study combed the empirical research on the influence of social interaction on online learning engagement,and analysed online learning engagement from the perspective of ecology,and also used the four dimensions(behavior,social interaction,emotion and cognitive engagement)to construct online learning engagement evaluation model,which provides theoretical support for online learning engagement analysis based on multi-source information.Secondly,the evaluation model based on BP neural network was proposed.Based on the analysis model of online learning engagement,this study constructed an analysis framework of online learning input and determines the evaluation index system of online learning input,combining with the general characteristics of students’online activities and log records.Then,an improved analytic hierarchy process(AHP)was proposed to determine the weight of each index,and on this basis,a BP neural network evaluation model was constructed.Finally,the results of NSSE-China learning engagement scale filled out by students were used as calibration,and the online learning engagement of students calculated by BP neural network model was compared with it.The experimental results showed that there is a significant correlation between the results calculated by BP neural network model and those calculated by NSSE-China scale.Thirdly,the evaluation model of emotional cognitive engagement was constructed.In order to solve the problem of inadequate representation of emotional and cognitive dimension information in learning engagement assessment model based on log data,an emotional recognition and engagement assessment model integrating multi-modal data was constructed,which used camera data and mouse movement data to complete the evaluation of emotional and cognitive engagement.In the training of the model,self-built training data sets were used to train the model,and multi-modal data were used to improve the accuracy of labeling.In the aspect of model building,an integrated model was constructed,which combines the expression recognition sub-model with the mouse data recognition sub-model to form a comprehensive recognition model.From the final test results of the model,the accuracy of the integrated model is higher than that of each sub-model.Fourthly,an evaluation and analysis method integrating multi-source data was constructed based on online learning engagement model.Based on the online learning engagement model proposed above,the evaluation results of affective cognitive engagement and the evaluation results of BP neural network based on log data were fused.In the fusion process,the modal data with different evaluation granularity were processed in two dimensions to realize the fusion analysis of different data.