Students’ Daily Behavior Analysis And Learning Situation Prediction Based On Campus Wi-Fi Data

Following the development of Internet technology and the improvement of information technology used in universities,universities are gradually ensuring a full Wi-Fi coverage on campus.Students and staff can use campus Wi-Fi to access the Internet for online learning and entertainment anytime and anywhere.Students’ daily behavior data can be obtained by collecting,processing and analyzing students’ wireless Internet access information,and through the analysis of students’ daily behavior data and prediction research based on students’ daily behavior,it can help schools manage students,increase campus learning atmosphere,and build people-oriented smart campus.This thesis analyzes and examines students’ behavior based on the data generated when they use Wi-Fi,and also proposes the idea of using the SSA-LightGBM model to predict students’ learning situation.The main studies of this thesis are as follows:(1)Based on campus Wi-Fi data,Baidu map API is used to visually analyze students’ daily behavior,and K-Means algorithm is used to perform cluster analysis of students’ daily behavior.First,the daily activities of students in the school,such as the location and trajectory,are displayed and analyzed using Baidu map API.The results show that Wi-Fi data can effectively show the daily behavior rules of individual and group students.Then,the behavioral characteristics of students are collected and characteristic vectors are formed.Students are divided into three types based on analyzing the proportion of time they spent in different places using the K-Means Clustering Algorithm: the studying type,the lacking-in-socializing type and the sports type.On this basis,further clustering analysis is carried out on different types of students according to gender and educational background,and it can also be obtained that the proportion of male students is less than the proportion of female students,and the proportion of undergraduate students is less than that of graduate students.Through the analysis of the clustering results,it is found that there is a correlation between students’ daily behavior and their learning enthusiasm.For example,students whose time distribution hotspots are concentrated in learning locations generally have higher learning enthusiasm.(2)This thesis argues that students’ learning situation can be predicted by using the SSA-LightGBM model.Since students’ daily behavior is relative to how active they are in learning,and different degrees of learning initiative influence students’ exam results to some extent,this thesis argues that students’ learning situation can be predicted by the SSA-LightGBM model.Parameter values are not easy to determine and have a noticeable influence on the performance of the model.Thus,the SSA algorithm is introduced to optimize the parameters.Comparative experiments are conducted by using SSA-LightGBM,LightGBM,random forest and support vector machine respectively.Results show that SSA-LightGBM has the highest prediction accuracy of 0.925.In terms of the time for prediction,SSA-LightGBM takes less time 33.47 s compared with random forest or support vector machine.Therefore,SSA-LightGBM performs well on both prediction accuracy and prediction efficiency when predicting students’ learning situation.The experiment in this thesis is based on the self-built student behavior data set.Through cluster analysis,the distribution law of students’ daily behavior is obtained,the SSA-LightGBM model is used to predict students’ learning situation.The experimental results show that the effect of the SSA-LightGBM model is better than that of other common models.The use of students’ Wi-Fi data can effectively predict students’ learning situation.By giving academic help to students at risk of failure,it helps to reduce students’ failure.rate and increase the learning atmosphere of the campus.