Mining Behavior Patterns Related To Academic Achievement Based On Mobile Sensing Data

Driven by the popularization of mobile phones as well as their enriching functions,nowadays we can infer users' daily behaviors using data collected from mobile phones.It's no doubt that how to improve academic performance is one of the most attractive topics among students.The impact factors,related behaviors and the prediction of grades have been paid a lot of attentions for a long time.This thesis focuses on how analyze the relationship between students' daily behaviors and their scholastic performance,using mobile sensing data.We first utilize an open mobile sensing dataset to analyze the relationships between GPA and daily behaviors.After that,we design and implement a mobile sensing data platform,collect mobile sensing data of Chinese students,and analyze the relationships between their GPAs,graduation whereabouts and behaviors.First of all,we propose a behavior inference model based on mobile sensing data,and infer some common behaviors of students,such as studying,attending class and physical exercise.We conduct correlation analysis between these behaviors and GPAs using an open mobile sensing dataset: StudentLife.The result shows that students who participate in more art activities are less likely to get high GPA,while there is no significant relationships between their GPAs and the time they spend in studying,attending class and discussing.After that,we present a heuristic method based on genetic algorithm to mine unknown behaviors that are related to GPA.We also use the StudentLife dataset to verify the algorithm.The results show that,late sleeping,staying in library but not focusing on study,spending time with roommates in dormitory are all adverse to the GPA of students,but spending time with many people often is beneficial to GPA.Finally,we design and implement a new mobile sensing data platform named StarLog,based on an open-sourced framework called UbiqLog.We deployed a data collection project in the university for 20 days,and eventually obtained 2.88 G data from 21 participants.We conduct data analytics on these data about the relationships between phone usage,mobility and GPA,graduation whereabouts of students.The results show that,the frequency of phone usage is negatively correlated to GPA,and mobility is not correlated to GPA.Besides,students who tend to continue their postgraduate study to get a doctoral degree spend much more time in using online shopping,chatting,sharing and browsing appliclations that those who tend to work after graduation.