Research On Learning Effect Prediction Method Oriented To Online Learning Behavior Analysis

The popularization of Internet technology and education informationization has promoted the rapid rise of online learning represented by MOOC classroom and made it widely recognized.The "Internet+education" proposed by China is based on the continuous development of science and technology.It is a new kind of learning environment with advanced education theory,rich curriculum resources,flexible teaching methods and diverse learning evaluation.The online learning platform stores a large amount of learner learning data.In order to explore the underlying laws contained in behavioral data,this paper studies the learning behavior of online learners in anticipation of improving the learning quality of learners.Therefore,it is of great practical significance to predict and intervene the learners' learning effects.However,by studying the literature,there are very few empirical studies in China.These studies either have one-sidedness on the data indicators,or the single prediction model used is not accurate.In view of the shortcomings of the summary,the main research contents of this paper are as follows:1.Combining learners' entire online learning process in VLE online learning platform,this paper proposes 33 data metrics for data analysis in 7 dimensions.These indicators not only include man-machine interaction learning behaviors such as content learning behavior,test behavior,cooperative behavior and resource search behavior,but also include indicators such as demographic background information,learning ability,learning attitude and other information that affect learning effect,so as to ensure learning influencing factors to the maximum extent.2.This paper conducts descriptive statistics on relevant data indicators of learners,and discusses whether online learning behavior really affects the learning effect.At the same time,six kinds of prediction algorithms,such as multivariate statistical method and machine learning algorithm,are used to model the learning effects.These algorithms are multiple regression algorithm,principal component regression,TAN algorithm,RBF-SVM,MLP algorithm and CHAID algorithm.Combining the training set and the test set,the performance of these six predictive models is demonstrated,and the advantages and disadvantages of these predictive models are distinguished.3.Aiming at the insufficiency of single prediction model,such as low precision,instability and inability to fully reflect the behavior index,this paper first puts forward the idea of mixing based on multiple single models,and based on this,gives an adaptive hybrid learning effect prediction.Model,which is the optimal combination of a single prediction model elected by the adaptive algorithm presented in this paper.It is then applied to empirical research and found that its accuracy rate is as high as 93%,which is 8 percentage points higher than the single prediction model.It is a good improved prediction model.Finally,based on the adaptive hybrid model,this paper proposes a prediction mechanism based on the whole process of online learning,draws a visual concept map and explains its role.