Study Of Students’ Performance Prediction And The Strategy Based On The Machine Learning

In the new period of digital transformation of Chinese education,when Chinese education informatization is moving towards intelligent education,online education has been booming,and various kinds of online and offline mixed teaching has become the new normal of education.However,in the online learning environment,due to the differences in learning ability,cognitive style,as well as the difficulty in unifying external conditions and other factors,learners have problems in the process of learning,such as poor results,low passing rate of course exams,and even drop out.At the same time,in blended teaching environment,teachers are often unable to supervise every learner,resulting in learners unable to timely and accurately obtain learning improvement strategies and other problems.Therefore,how to use information technology to accurately predict the phased learning effect of students in the teaching process,and according to the predicted results to timely intervene the poor learning effect of students and provide learning performance improvement strategies,will become an important way to improve the quality of online education and hybrid education.Based on this,this paper,aiming at two different scenarios of online learning and blended learning,carries out research on students’ learning achievement prediction model and achievement improvement strategy based on machine learning Algorithm Light GBM and Genetic Algorithm(GA).The main work includes the following aspects:(1)Research on the prediction model of student learning performance.Aiming at the lack of interpretability of the existing methods in the task of performance prediction,this paper constructs a student grade prediction model based on Light GBM,and uses Grid Search CV function to determine the optimal parameters of the model.Through the MOOC dataset of OULAD(Open University Learning Analytics Dataset)and the hybrid "Gold Course" dataset of University Computer Foundation,This paper verifies the validity and accuracy of the proposed model in predicting student achievement in various teaching scenarios,and analyzes the importance of characteristics of two classes of courses based on the interpretability of Light GBM algorithm.(2)Research on the Prediction Model of student Learning Achievement.Aiming at the lack of interpretability of the existing methods in the task of perfoamance prediction,this paper constructs a student performance prediction model based on Light GBM,and uses Grid Search CV function to determine the optimal parameters of the model.Through the MOOC dataset of OULAD(Open University Learning Analytics Dataset)and the hybrid "Gold Course" dataset of University Computer Foundation,This paper verifies the validity and accuracy of the proposed model in predicting student achievement in various teaching scenarios,and analyzes the importance of characteristics of two classes of courses based on the interpretability of Light GBM algorithm.(3)Study on strategies for improving students’ academic performance.Based on the results of student perfoamance prediction,this paper uses Genetic Algorithm to construct the strategy model of student achievement personalized improvement.Given the students to be improved,the model encodes their optimizable learning behavior characteristics into the corresponding multiple chromosomes.Under the guidance of the fitness function,the model optimizes the strategy through the single point crossover,mutation and other operations in the genetic algorithm,so as to search for the most suitable achievement improvement strategy for the students.This paper verifies the effectiveness of the promotion strategy in two teaching scenarios.