A Study On Academic Performance Analysis And Prediction Based On Education Big Data

The application of Internet in the field of education and teaching is increasingly extensive,and a large amount of educational data is generated in this process.How to use these massive educational data reasonably is always an important topic in the field of educational data mining.A student’s daily behavior can have a great impact on his academic performance.To predict students’ academic performance based on their daily behavior data,certain interference measures can be taken to the students with poor academic performance risk in the early stage,so as to avoid the occurrence of students’ study and life affected by poor academic performance to the greatest extent.The existing researches on students’ academic performance mostly use the methods of questionnaire survey and prediction of students’ course scores,but the data used by these methods have certain limitations.Some traditional machine learning models,such as SVM and logistic regression,are often used to predict the academic performance of college students.However,these methods cannot make use of the information from the time series data of students’ behavior to predict.In addition,the use of complex models on simple student achievement data sets can lead to over-fitting problems.Therefore,this paper proposes a prediction model of academic performance based on the multi-dimensional characteristics of college students,which can analyze and predict the academic performance of college students in a targeted way,detect students with poor academic performance in advance and carry out intervention work.The work of this paper includes:(1)In view of the data limitations,this paper analyzes students’ behaviors at school from the three dimensions of students’ basic information data,students’ performance data and students’ behavior data,constructs behavior entropy,which is used to mine the regularity of behavior characteristics and accurately describe the behavior laws of students(2)In view of the problem that a large number of invalid rules will be generated when the traditional association rule mining algorithm is used to mine the behavioral data of college students,an improved association analysis algorithm is proposed to analyze the correlation between students’ various behaviors and mine the behaviors with strong correlation with students’ academic performance.(3)In view of the traditional model can’t better capture the behavior of time-series data,this paper proposes a causal convolution based network and the network structure of residual structure,personalized analysis and prediction model of students’ academic performance,to predict the students with poor academic performance risk early intervention,further improve students’ academic performance.