College Students's Academic Warning Analysis Based On Big Data

In recent years,with the repaid development of internet technology,big data becomes the focus of all trades gradually.How to use education big data to enhance education effects and improve the quality of talents are the big issues of education field.For this problem article proposes score alert analysis bases on students behavior big data to supply decision –making reference for university manager.Firstly,this article acquires college students behavior data from on-line transactionprocessing systems in 2017,this article extracts the original data information into variables which can be used for analyzing and researching,in next step this article screens the variables preliminarily and deals the missing values as well as outliers;after that according to the big number of variables the article adopts principal component analysis and principal basement analysis two dimensionality reduction methods,to form new variable sets;and separate examples into training set and testing set by 9:1;Bases on these two new sets the article does regression analysis to training set respectively,through evaluation and comparation,the article chooses the variable set from principal basement analysis to be independent variable set for the regression model eventually,According to final model the article predicts that this college will have 448 students been left behind on academic scores in next semester and advises college to alert these students on academy,at the same time through research prediction model`s variables`coefficients and the correlativity of variables,the article finds the students with high ratio of weekend outgoing,more times of self studying,normal family financial situation,preference for browsing academy and job websites will have better academic performance and advises colleges to lead students to such fields.The article overcomes the shortcoming of traditional research that only use single factor such as achievements of various subjects and levels of teachers to research students academic achievements,and focuses on many factors in students behavior which can affect academic performance.But because of the restrictions on data,this article exists some flaws on indexes choose and research directions which can be improved in future research.