Research On Analysis And Prediction Of Online Learning Behavior

The rapid advancement of Internet technology has promoted the rapid development of the education industry,and the education industry has entered the Internet age.Although online learning provides learners with a free and convenient way of learning,there are still many problems behind them.The low pass rate and high dropout rate are the key problems to be solved.Aiming at these bottleneck problems in the development of E-learning,this paper explores and analyzes a large number of learning behavior data generated by learners in the process of E-learning,and uses ensemble learning algorithms to classify learners.Implement different teaching programs according to different types of learners,thus providing a scientific basis for efficient teaching of online learning platforms.This paper uses the ed X network learning dataset published by the Massachusetts Institute of Technology and Harvard University to conduct learner classification research.Firstly,the characteristics of learning behavior in ed X dataset are analyzed in detail,and some conclusions are obtained by using statistical analysis method.At the same time,some methods of feature engineering are used to process the dataset accordingly in preparation for the subsequent construction of the classification model.Secondly,in the aspect of classification model construction,through the detailed study of the theory of Stacking algorithm,according to its performance advantages,it is using as the learner category prediction algorithm.According to the characteristics of the Stacking algorithm,it is improving in terms of hierarchical structure,data feature attributes representation,combination strategy and classification algorithms.The improved Stacking algorithm is using to construct the classification model.Finally,the improved Stacking algorithm was tested by using 15 different types of UCI datasets.The results show that the improved Stacking algorithm has better performance in accuracy,precision and 1 values.The improved Stacking algorithm and the unimproved Stacking algorithm are used to classify and predict the public dataset of the ed X network learning platform.The experimental results show that the improved stacking algorithm improves the accuracy,precision and 1 value of the ed X dataset by 2.46%,2.11%,1.97%,thus verifying the practical effectiveness of the improved Stacking algorithm for Online learners' classfication.