Research On Student Performance Prediction Based On Campus Big Data

With the rapid development of information construction in Colleges and universities,the information system represented by the campus all-in-one card system has been widely applied to all aspects of students’ life,and a large amount of student behavior data has been generated.How to mine useful information from these data and analyze the correlation model between students’ behavior and achievement is a hot research topic in the current education field.This paper is mainly based on the performance prediction research of student behavior,combined with data mining classification algorithm and deep learning algorithm to analyze the behavior data of students in campus activities,trying to find a high accuracy prediction method of students’ performance.Firstly,the behavioral data such as consumption data,borrowing data,study duration and library access control are analyzed by clustering and association rules,and the behavioral patterns hidden behind students’ behaviors and the influencing factors related to students’ performance are mined;Secondly,a performance prediction fusion model based on random forest,gradient boosting decision Tree(GBDT)and extreme gradient boosting(Xgboost)is established;Finally,aiming at the problems of low prediction accuracy of the fusion model and insufficient manual extraction of behavioral features,a CNNLSTM performance prediction model based on attention mechanism is proposed.The main research contents of this paper are as follows:(1)Data preprocessing and feature extraction.Aiming at the problem that the current student information management platform is not perfect and the effect of mining useful information is not good,a combination algorithm of student behavior analysis based on clustering and association rules is proposed.Firstly,K-mediods(Kcenter point)algorithm is used to cluster the student behavior data,and the clustering results are discretized;Then,Eclat algorithm is used to analyze the correlation between student behavior and performance.Finally,combined with the two algorithms to comprehensively analyze the behavioral factors that affect student performance(2)Construct a performance prediction model based on data mining algorithms.According to the traditional manually extracted behavior features,a performance prediction fusion model based on random forest,GBDT and Xgboost is established.Firstly,the weights of the single classification models of random forest,GBDT and Xgboost are calculated by the Boosting algorithm,and then the above single models are fused by weighted average method.The fusion model is compared with the above single models to verify the effectiveness of the fusion model.(3)Established a CNN-LSTM student performance prediction model based on the attention mechanism.On the basis of the fusion model analysis results,a CNN-LSTM student performance prediction model based on attention mechanism is proposed.Firstly,convolutional neural network(CNN)is used to extract deep students’ behavior features,and the maximum pooling method is used to select the significant features of students’ behavior features;Then,the extracted features are used as the input of long and short-term memory network(LSTM)to predict students’ performance;Finally,the time-series attention mechanism is introduced at the output of LSTM to allocate attention weight for students’ behavior features in different weeks,and select the key point data,so as to predict student performance more accurately.