Research And Application Of Online Course Student Behavior Evaluation Model

Online courses are divided into practical courses and theoretical courses according to the different learning styles of students,and practical courses are usually carried out in engineering practice in groups.In the evaluation of practical courses and theoretical courses,due to the lack of supervision of teachers and schools on the online platform,irregular behaviors such as ghostwriting are inevitable.At the same time,the evaluation process lacks the specific application of data in the learning process of students’ online courses,which affects the accuracy of the final results.Therefore,there is an urgent need to use learning process data to evaluate students’ learning behavior in online courses in order to improve the effectiveness and objectivity of students’ comprehensive performance evaluation.While predicting student performance in online courses,the current approach relies heavily on students’ past academic performance,disregarding their current learning behavior.This approach overlooks critical factors that can impact students’ success in the current course stage,making it essential to develop an effective classification algorithm.To address the challenges of evaluating student behavior and predicting their performance in online courses,this thesis proposes two innovative models: Group Engineering Practice Prediction Framework based on Bidirectional Long Short-Term Memory network(GEPBi LSTM)for engineering practice courses,and Three-layer Ensemble Learning Framework for Predicting Student Performance of Online Courses(TELF-PSPOC)for theoretical courses.The main contributions of these two models are as follows:(1)GEP-Bi LSTM framework: Extracting features of each stage from the groups,organizing the practical situations of each group.Constructing group vectors for the practical process of each group,providing a comprehensive description of the completion status and specific content of group practices.The bidirectional long short-term memory(Bi LSTM)network combined with attention mechanism can capture information from both forward and backward data in the sequence,improving data utilization.(2)The feature enhancement method for online course student behavior: The TELFPSPOC model adopts a feature enhancement method that considers the behavioral characteristics of students in online courses,such as their scores and pass rates of all stages of tests,daily hits of online resources,and more.(3)The Three-Layer Ensemble Feature Learning with Heterogeneous Classifiers(TEFLHC): Based on a two-layer stacking ensemble model,TEFL-HC integrates balanced tree models and neural networks,consisting of a pre-training layer,a base learner layer,and a meta-learner layer.Its distinguishing feature is that the model is capable of extracting students’ performance over time while retaining interpretability,which enables educational researchers to gain indepth understanding of the learning process.Furthermore,this thesis has developed an online course student behavior evaluation system that analyzes the learning process data and evaluates the learning behavior.This system aims to gives early warning to the students who cannot pass the course judged by the system,and feeds back the results to the teacher.