Learning Behavior Analysis And Academic Prediction Research Based On Online Course Data

In the context of the era of big data,with the development of information technology,intelligent education as an innovative application of artificial intelligence has accelerated,and intelligent learning environment has also become the mainstream trend of learning.The learning behavior data generated by learners when they perform online learning or learningrelated activities can to some extent show the learner’s preference tendencies in the learning process,that is,learning style.With the help of information technology,mining the potential characteristic information behind behavior data and conducting research and analysis can provide decision support for both teaching parties and managers to choose learning resources and methods,conduct accurate teaching and improve teaching quality.Therefore,the research purpose of this paper is to reveal the learning style in online learning behavior,analyze behavioral characteristics and predict learning outcomes through technical means.The topic of this paper is to classify learning styles and build performance prediction models.Combining the theories of behavioral science,online learning and learning analysis,this paper studies and analyzes the relevant theories and experimental results from the aspects of the connotation of online learning style,behavioral performance,and academic prediction modeling.The research work of this paper mainly includes the following aspects:（1）This paper first expounds the background of the topic selection of the paper,and discusses the practical significance of mining online learning behavior data,uses the literature review method to sort out the research status at home and abroad from the research content and research methods of online learning behavior analysis,and proposes the research problems of this research according to the existing research problems and the actual situation,and proposes research ideas and research methods based on the research problems.（2）This section expounds the relevant theories of learning behavior analysis and the techniques and methods related to data mining,including behavioral science theory,online learning related theory,learning analysis related theory,and data mining definition,process and common algorithm.（3）On the basis of studying the theoretical model of learning style,this study deeply understands the connotation of learning style,and combines the characteristics of online learning,and constructs an online learning style model based on the Four Dimensions of Information Input,Information Processing,Learning Attitude and Learning Behavior Input based on the Felder-Silverman model.This section divides the learning style through the KMeans algorithm,analyzes the differences in the learning style and learning behavior of learners in various dimensions,and establishes a student learning profile.（4）In order to further explore whether learning style affects learning performance,this paper uses ANOVA and independent sample T test,and through the analysis results,it can be seen that the performance of learners with different learning styles has significant differences.This paper introduces data preprocessing and feature selection,and combines machine learning classification algorithms to build a learning effect prediction model based on online learning behavior analysis.（5）In this paper,the "University Physics Experiment" course on the superstar platform applies the learning effect prediction model to predict and analyze the learning effect.Based on correlation analysis and recursive feature elimination,five learning behavior features with great influence on grades are selected,and a prediction model for academic performance is established based on five algorithms in machine learning.By using the grid search method for parameter optimization,combined with the advantages and disadvantages of five prediction models,the model performance is demonstrated on the test set.Experimental results show that the prediction model after feature selection and parameter optimization can effectively identify students’ behavioral characteristics and predict students’ achievements to a certain extent.Experimental comparison and analysis of the prediction effects of five machine learning prediction algorithms show that Ada Boost classification is the best.