Analysis And Identification Of Students’ Risk Of Internet Addiction Based On Educational Big Data

The rapid development of the Internet has brought great convenience to people’s lives,but the excessive use and dependence of the Internet will cause the risk of Internet addiction.Internet addiction not only affects students’ study and life,but can even lead to more serious consequences such as depression and suicide.Therefore,it is very necessary to detect high-risk groups of Internet addiction in advance,and to intervene and treat in time in the early stage.At this stage,most of the detection methods for Internet addiction use psychologist questionnaires,but this method has certain limitations.At present,most studies in the computer field use traditional machine learning algorithms such as naive Bayes and logistic regression for modeling,and these methods cannot capture the timing information in the data.In addition,using complex models on simple data sets can easily lead to overfitting problems.Therefore,this paper establishes a personalized early warning mechanism that uses time series information in educational big data to detect early symptoms of Internet addiction.Specific work includes:(1)Aiming at the limitations of using questionnaires,this article is based on educational big data,constructing students’ temporal characteristics in five different dimensions for early symptom detection of Internet addiction,and assessing students with high risk of Internet addiction and students with low risk of Internet addiction.Analysis of differences can detect high-risk groups of Internet addiction in advance,obtain the characteristics of early symptoms of Internet addiction,remind schools and parents to pay attention to these students in time,and take necessary measures to reduce the adverse effects of Internet addiction.(2)Aiming at the problem of insufficient capture of time series information by traditional machine learning algorithms,this paper constructs a prediction model based on LSTM and adds an Attention mechanism to enhance the impact of important features on Internet addiction.By comparing with the baseline model and comparing the time series features Analysis of ablation experiments,the experimental results can fully illustrate the effectiveness of the model,and can better capture the timing characteristics in the data.(3)Aiming at the model over-fitting problem caused by the use of complex models on simple data sets,this paper proposes the gd-LSTM algorithm,which uses the dropout principle to make the three gates of LSTM randomly do not play a role with a certain probability.Realize the LSTM cyclic layer regularization technology to solve the problem of model overfitting.The experimental results show the effectiveness of this method and can better solve the over-fitting problem.This paper establishes an automated personalized early warning mechanism for detecting early symptoms of Internet addiction,which can remind schools,teachers or parents to take necessary measures in time to reduce the harm caused by Internet addiction.The work in this article also helps psychologists discover the patterns of Internet addiction,and extract the factors that cause Internet addiction from problems that seem to have nothing to do with Internet addiction.