Research On The Intelligent Evaluation Method Of E-learning Based On Deep Learning

The development of Internet applications has caused profound changes in the field of education,especially since the "14th Five-Year Plan",with the impetus of the epidemic,E-Learning has become an indispensable part of school teaching.However,the current E-Learning is faced with problems such as the learning process cannot be effectively supervised,the learning enthusiasm is difficult to judge,the credibility of the assessment results is low,and the evaluation method is mainly the final evaluation based on the examination,and the lack of multi-dimensional formative evaluation.To this end,the relevant design and evaluation methods have been studied in order to clarify the evaluation path and achieve an objective and accurate evaluation of E-Learning.The emotional tendency,learning status and learning effect of E-Learning are the three important evaluation dimensions of E-Learning evaluation,these three-dimensional evaluation is the key to adjusting learning enthusiasm,improving learning status,and increasing the objective credibility of learning certification.Therefore,the online learning evaluation process is designed,which consists of data acquisition layer,intelligent evaluation layer and result application layer.The data acquisition layer analyzes the types and acquisition methods that can be obtained by the current E-Learning platform;the intelligent evaluation layer introduces the basic process of intelligent evaluation model construction including E-learning sentiment analysis and learning state prediction;and the results application layer detailed how the evaluation results of the three aspects of emotional tendency,learning state and learning effect are applied to formative assessment and summative assessment.In view of the implementation of the three evaluation dimensions in E-Learning evaluation process,deep learning-based method was studied by using the existing collectable data.For sentiment analysis,the learner’s emotional tendency is judged by integrating the attention mechanism into Text CNN to improve the model;For the learning state,the learning behavior data is enhanced by using student information and course information,and the information extraction is carried out by Convolutional Neural Networks and Gate Recurrent Unit and the characteristic weighting of the Self-Attention mechanism is used to realize the timely prediction of the learning state of students;For learning effect,the prediction result of the learning status is taken as the weight of the classroom assessment score,and the comprehensive in-class test and stage evaluation as the final course results to achieve the evaluation of the learning effect.The effectiveness of the improved Text CNN-based sentiment analysis model is verified through the real learning review data in the MOOC platform of Chinese universities,and the results show that the improved method performs better on F1 than other artificial neural network methods.Experiments on two public datasets,MOOCCube and Xuetang X,verify that the proposed E-Learning state prediction model based on data enhancement and feature weighting can predict the learning state of learners more timely and accurately,and has good robustness.