Study On Prediction Model Of Online Learners' Achievement Based On Learning Analytics

As the size of online learners increases,academic warning of online learning has become a difficult teaching problem for learners.Based on the learning analysis technology and the data obtained from the online learning platform,this paper attempts to establish a model that can predict academic performance to optimize academic early warning and teaching strategies.(1)Firstly,the difference between learning analytics and educational data mining and the connotation of learning analytics are analyzed.The learning analytics process and element model are proposed,and the links and components of learning analytics in the teaching process are expounded.Then the "learning" element framework is proposed,and "learning" is described in many aspects.Based on the "learning" element framework,the paper analyzes the impact of learning behavior input on academic achievement,and proposes a learning behavior input evaluation framework to measure the degree of learning behavior input from multiple dimensions.(2)Using decision tree and random forest algorithm as the implementation method of learning achievement prediction model.First,the algorithm principle is analyzed.Then the algorithmic idea of information entropy is used to discretize the continuous variables in the data set.Experiments show that the method can effectively improve the fitting degree of the algorithm.Then the algorithm is used to model the dataset,and the prediction model with good effect(accuracy rate of 70%)is obtained by means of parameter adjustment and pruning.Finally,the model is evaluated by the confusion matrix.The accuracy of the model for the sample prediction classification is H(84%),M(69%)and L(89%),indicating that the model is effective for academic warning.(3)The prediction results show that the “student absenteeism” and the four characteristics describing the learning behavior(viewing the number of resources,number of raising the hand,number of viewing announcements and number of discussions)have the greatest impact on academic performance.Through the correlation analytics of the four characteristics of learning behavior,the “discussion number” is lower than the other three correlations,and the other characteristics are strongly correlated with each other.The result is helpful for learning behavior input evaluation framework.Dimension division.