Research On Achievement Prediction Model Based On Online Learning Behavior Analysis

With the development of big data and education informatization,the use scale of online learning platform is expanding day by day.In the process of interaction between learners and online platform,a large amount of behavioral data is generated.By mining and analyzing these behavioral data,we can better understand the online learning situation of students and find out the learning rules of students,etc.,.This will help students in the process of learning real-time intervention,targeted guidance,so as to achieve the goal of personalized training.Therefore,it is of practical significance and academic value to use the learning behavior data generated by the online learning platform to predict students' academic performance and achieve the purpose of intervening students' academic performance in advance.The research work of this paper is mainly in the following three aspects:Firstly,this study collected the online learning behavior data of students through the online learning platform,and further explored the influence mechanism of online learning behavior characteristics and academic performance with the help of visual analysis methods(scatter diagram,box diagram,etc.),so as to analyze the characteristics of different learning behavior characteristics in different achievement ranges.At the same time,based on the important feature selection algorithm of random forest,this paper is used to select the online learning behavior features that have more impact on students' achievement,and analyzes the difference between them and the online behavior features of this research field,which lays a foundation for the subsequent construction of achievement prediction model.Secondly,LightGBM algorithm has the characteristics of high accuracy and strong generalization ability on the classification problem,and has great advantages in noise processing and distributed processing.In this paper,LightGBM algorithm is used to constructed a learning performance prediction model based on the selected learning behavior characteristic data,and then the network search algorithm is used to optimize the model parameters to improve the overall performance of the model.Finally,this paper proposes a performance prediction model based on SMOTE over-sampling and Lightt GBM.Aiming at the problem of low recognition of a few samples in Lightgbm-based performance prediction model,which leads to low overall prediction effect of the model,SMOTE over-sampling method is adopted to optimize the performance prediction model to effectively solve the problem of data imbalance and improve the accuracy of performance prediction.The following conclusions are drawn in this study: 1.Analysis of feature selection results based on random forest shows that chapter learning times have the greatest impact on grades,followed by video viewing time,chapter average score,task point completion rate,etc.Therefore,in the process of online learning,teachers need to focus on these learning behaviors of students,so as to guide students in learning more effectively.2.LightGBM algorithm has a good application effect in the evaluation of performance prediction model,indicating that the performance prediction model using this algorithm can effectively identify the features of high-dimensional learning behaviors.3.The performance prediction method based on SMOTE over-sampling and LightGBM can improve the accuracy,recall rate and F1-score of the model to a certain extent.