Augmented: Academic Performance Prediction Based on Digital Campus

Abstract

Digital data trails from disparate sources that cover different aspects of student life are being stored daily on most modern university campus. However currently it remains a challenge to (i) combine these data into a holistic view of a student; (ii) use this to predict academic performance accurately; (iii) and take advantage of the prediction to drive behavioral change and support student positive engagement with the university. To initially alleviate this problem, a framework named Augmented Education (AugmentED) is proposed in this chapter. In our study, (1) firstly, the experiment is conducted based on a real-world campus dataset from college students (N = 156), aggregating multisource data that consists of a small private online courses (SPOC) platform, usage of smart cards (e.g., library entry, meal, and consumption), Wi-Fi detection records, and central storage (e.g., gender, age, and class schedule). Specially, to gain a deep insight into the features leading to excellent or bad performance, on the one hand, three novel metrics (e.g., Lyapunov exponent) that measure the regularity of campus lifestyles are estimated; on the other hand, LSTM-based features that represent the dynamic changes of temporal lifestyle patterns are extracted by means of long short-term memory (LSTM). (2) Secondly, a machine learning-based intelligent algorithm is developed to predict academic performance. (3) Finally, visualized feedback is designed, potentially enabling students (especially for at-risk students) to optimize their interactions with the university and achieve study-life balance.