Research On Student Performance Prediction Method Via Online Learning Behavior Analytics

With the continuous development of online learning platforms,educational data analytics and prediction has become a promising research field,which is helpful to the development of personalized learning system.However,the study of domestic learning analysis and prediction model still stays in the stage of theoretical exploration and construction,Just a very small number of studies actually do the prediction by the online learning platform data,but when choosing the prediction algorithm,there is no description of the selection basis,which will cause blindness in the selection of the prediction algorithm for follow-up studies.Looking at the research at home and abroad,when selecting behavior indicators for the predictive model,it does not combine the whole online learning process.Most studies directly select behavioral indicators,which inevitably results in the absence of behavioral factors,and thus affect the prediction results.The main research contents and research results of this paper are as follows:1.This thesis summarizes and analyzes relevant researches on learning analysis models and learning predictions at home and abroad by the literature research method.Based on the research results summarized in this study,combining with the actual situation of the Hstar in this study,the main issues of the research are proposed.2.Combining the whole online learning process,clarifying the behavioral indicators that influence the learning effect in related literature,on the basis of this,we proposed 19 behavioral indicators for data collection,and constructed a learning effect prediction model based on online learning behavior analytics.The model consisted of four parts:data collection and pre-processing,learning data analytics,and learning effect prediction algorithm,prediction and intervention,the prediction model training included four parts:feature value selection,prediction algorithm selection,data set classification,and prediction model evaluation.3.Taking a course in the Hstar at CCNU as a case,an experimental study based on the learning effect prediction model constructed in this paper was conducted.The experimental study was conducted in two parts,Firstly,according to the demand of high accuracy,we used Logistic regression algorithm to predict the student performance,the experimental results showed that all of the evaluation index of the prediction model was up to 90%,Explained that the model can predict student performance very well.Secondly,according to the demand of high explanation,we used ICRM evoluted from genetic programming algorithm to predict student performance.By comparing with other interpretable prediction algorithms,ICRM algorithm has a stronger degree of explanation in ensuring competitive prediction accuracy.