Research On Asssessment And Prediction Method Of E-Learning Learning Behavior Based On Sentiment Analysis

As an important part of Chinese Educational Informatization,E-learning teaching mode,with its unique open education characteristics,breaks the time and space constraints of traditional teaching mode,boosts the sharing of high-quality educational resources,and promotes the popularization and individualization of education.Undoubtedly,this teaching mode has achieved great success in the reform of teaching methods.However,problems such as the difficulty in handling big data,high registration rate and low graduation rate have become an important problem restricting the sustainable development of this teaching mode.How to effectively analyze and evaluate students’ learning behavior,so as to timely discover their dropping out tendency and take targeted teaching intervention is the key problem to improve the core competitiveness of E-learning teaching mode.For the four problems that E-learning education platform is difficult to effectively handle large data,E-learning learners’ learning behavior is difficult to accurately quantify evaluation,E-learning learners dropped out of school is hard to find in time,and E-learning learning platform can’t take targeted teaching intervention,this thesis takes E-learning learners’ text data,graduations,and other learning behavior as the research object,propose E-learning learning behavior assessment and prediction methods based on sentiment analysis.Firstly,for the problems that the unstructured data of E-learning education big data is hard to effectively processing and the low granularity of sentiment analysis results in traditional emotion analysis methods,this thesis proposed a fine-grained-based multipolarity sentiment assessment model.This model based on the affective lexicon analysis method,and the positive and negative two polarized sentiments in the traditional sentiment analysis method were refined into eight basic human sentiments,including happiness,sadness,anger,fear,trust,disgust,surprise and expectation,so as to improve the granularity of sentiment analysis.Meanwhile,the model combine with the characteristics of Chinese emotional expression and semantic relation,established a relatively comprehensive sentiment quantitative rules,improving the accuracy of the affect intensity calculation.In addition,in order to further understand the trend of learners’ emotional changes,the dominant sentiments of learners in each stage were extracted,so as to construct the multipolarity emotional change chain of E-learning learners.Secondly,this thesis proposed a sentiment-change-trend-based behavioral prediction and course recommendation algorithm.This algorithm takes the mapping relationship between learners’ multipolarity sentiment change chain and graduation situationas as the research object,established the multivariate linear regression equation.Based on the analysis of the limitations of the min-batch gradient descent method,the improved learning rate hot start method was used to optimize its performance,and the multivariate linear regression equation was solved by the improved min-batch gradient descent method.In addition,the algorithm recommend courses of interest for learners with low probabilities of graduation or significant decline,and conditional probability and text mutual information evaluation are used as the screening criteria,with the attenuation effect of learning interest is taken into account,so as to improve their probabilities of graduation.Finally,the method in this thesis is verified by comparison analysis,receiver operating characteristic curve analysis,5 fold cross validation and other evaluation methods.The proposed method provides a new research idea for assessment and prediction of E-learning learning behavior,and learners’ tendency of dropping out is able to be discovered in time,with recommend the related interest courses to raise their graduation rates.On the other hand,the proposed method is able to optimize the learner’s learning experience,promote E-learning teaching platform development in personalized education and effective teaching,so as to provide Chinese Educational Informatization Reform with a certain reference value.