Predicting Academic Performance for College Students

Abstract

Detecting abnormal behaviors of students in time and providing personalized intervention and guidance at the early stage is important in educational management. Academic performance prediction is an important building block to enabling this pre-intervention and guidance. Most of the previous studies are based on questionnaire surveys and self-reports, which suffer from small sample size and social desirability bias. In this paper, we collect longitudinal behavioral data from 6, 597 students' smart cards and propose three major types of discriminative behavioral factors, diligence, orderliness, and sleep patterns. Empirical analysis demonstrates these behavioral factors are strongly correlated with academic performance. Furthermore, motivated by social influence theory, we analyze the correlation between each student's academic performance with his/her behaviorally similar students'. Statistical tests indicate this correlation is significant. Based on these factors, we further build a multi-task predictive framework based on a learning-to-rank algorithm for academic performance prediction. This framework captures inter-semester correlation, inter-major correlation and integrates student similarity to predict students' academic performance. The experiments on a large-scale realworld dataset show the effectiveness of our methods for predicting academic performance and the effectiveness of proposed behavioral factors. :2 H. Yao et al. students' academic performance. It has been demonstrated that physical status [23, 29], intelligence quotient [9] and even socioeconomic status [38] are correlated with academic performance. However, these characteristics are relatively stable over the long run and are difficult to change via educational management.Comparatively, more studies are focused on the perspectives of psychology and behavior, partially due to the possibility of intervening on student's mentation and behavior. Extensive experiments about the correlation between the big-five personality traits (Openness, Conscientiousness, Extraversion, Agreeableness, Neuroticism) and academic performance have been reported [2,25,33], uncovering conscientiousness as one of the strongest predictors. Behaviors like class attendance [8], lifestyle [34] and sleep habit [10] are also highly associated with academic performance. However, almost all of these results are obtained from questionnaires or self-reports, which usually suffer from small sample size and social desirability bias, resulting in the difficulty to draw a valid and solid conclusion.Thanks to the development of information technology, there is a growing trend to augment physical facilities with sensing, computing and communication capabilities in modern universities. These facilities provide an unprecedented opportunity to collect real-time digital records about students' campus activities in an unobtrusive way and to reveal behavioral predictors for academic performance. In this paper, mainly through campus smart card, we collect 6, 597 students' longitudinal behavio...

Behavior-driven Student Performance Prediction with Tri-branch Convolutional Neural Network

“…A detailed ablation study on various configurations of our approach is provided. In comparison with the baseline methods presented in [2,13], we demonstrate the potential of our approach for student performance prediction.…”

“…For student performance prediction, we seek to obtain a mapping function f : X → R, which yields a real value f (X ) indicating the performance grade of each student. Following the previous studies [2,13], we concentrate on the academic performance ranking between different students, and cast the problem of student performance prediction as a pairwise ranking task. To this end, we construct the ground-truth pairwise comparison by a triple…”

“…We compared the Tri-branch CNN against the baseline methods presented in [2,13], where three types of high-level behavioral characteristics of students are manually designed and extracted from their campus smart card records, i,e., orderliness, diligence, and sleep pattern. With each type of the handcrafted behavioral characteristics, we used the well-known learning-to-rank algorithm RankNet [1] to predict the academic performance rank of students.…”

“…For example, if a student spends a long time in the library every day, it is more likely that the student can achieve better performance in the exam. From this point, recent studies [2,13] manually defined and extracted several behavioral characteristics of students from their campus smart card records, such as orderliness, diligence, and sleep pattern, and used them to facilitate student performance prediction. However, designing effective student behavioral features by hand still requires a considerable amount of domain expertise and engineering skills.…”

Academic Performance Prediction Based on Multisource, Multifeature Behavioral Data

“…• Lifestyle Behaviors (e.g., eating, physical activity, sleep patterns, social tie, and time management) ; and • Learning Behaviors (e.g., class attendance, study duration, library entry, and online learning) ( [7,8,[23][24][25][26][28][29][30][31][32][33][34][35][36][37][38]). For example, [2] investigated the incremental validity of the Big Five personality traits in predicting college GPA.…”