Academic Early Warning Based On Students' Behavior Analysis

With the advent of the "big data" era,a series of concepts combining pedagogy,statistics,and computer science,such as educational big data,smart education,and educational data mining,have been proposed one after another.Governments,enterprises,schools,researchers,managers,teachers,the public,etc.are beginning to pay attention to education big data,and relevant policy documents,research institutions,academic activities,market products,etc.have begun to appear.As far as universities are concerned,on the one hand,universities attach great importance to the important role of information technology in school education,and at the same time actively integrate the school's data management system to make full use of existing data to serve teachers and students.However,given the complexity and large amount of education data,how to find valuable information from it and use it properly becomes difficult.On the other hand,the development of higher education from elite education to popular education has led to a significant increase in the number of college students,a gradual decline in the quality of students,and an increase in the number of students who are unable to graduate successfully.At the same time,counselors and class teachers cannot take care of students' thoughts,lives and studies,and it is impossible for them to understand the latest thoughts and academic difficulties of the students in time and accurately.Using data mining technology to find and solve students' academic problems and avoid students from repeating grades and dropping out is a problem worth studying in colleges and universities at present.This article first uses campus card data to analyze students' daily behaviors.Through the analysis and refinement of student consumption data and library access control data,indicators such as early rise index and library study time are extracted.Through descriptive statistics and visual analysis of student behavior indicators,grasp the overall behavioral characteristics of students;then combine the basic information of students and relevant information of the curriculum,analyze the impact of various indicators on student academic performance and the impact on changes in performance,and establish an index database;then students are divided into three categories:no-warning,first-level warning,and second-level warning according to the object of the early warning.The best warning combination is analyzed step by step based on historical performance warnings and student behavior warnings:historical warning levels combined with behavioral data prediction based warning level changes;five machine learning classification algorithms are used to predict student warning levels,and then the genetic algorithm is used to optimize the five classification models.The improved model has a correct rate of 95.52%.At the end of the article,it analyzes the shortcomings of the current university student's academic early warning mechanism and proposes an improved mechanism based on the obtained early warning model.The list of students to be warned are determined in advance through the academic early warning model based on student behavior data,so that these students can adjust in time before academic problems occur,thereby reducing the probability of academic problems and helping students complete their studies smoothly.