A MULTISOURCE MULTI FEATUREAPPROACH TO PREDICTING ACADEMIC PERFORMANCE
Abstract
On utmost ultramodern university premises , digital data trails from numerous sources that cover colorful angles of pupil life are kept on a regular base. Yet, it's still delicate to( i) combine these data to get a comprehensive picture of a pupil,( ii) use these data to prognosticate academic achievement effectively, and( iii) use similar prognostications to encourage positive engagement of scholars with the university. This paper suggests the Augmented Education paradigm as a first result to this issue. In our study, an trial is first carried out using a real- world lot dataset of council scholars, which collects multisource behavioural data encompassing both outside and outside the classroom as well as online and offline learning. In particular, criteria measuring the direct and nonlinear behavioural changes of lot cultures are estimated, and features representing dynamic changes in temporal life patterns are uprooted by the use of long short- term memory in order to gain in- depth sapience into the factors impacting excellent or poor performance. Second, bracket algorithms grounded on machine literacy are created to read academic success. The design of graphical feedback allows scholars to potentially ameliorate their connections with the university and strike a balance between their particular and academic lives. The trials demonstrate that the stoked model is able of soothsaying scholars' academic success with high delicacy.
Keywords
Behavioural pattern, Machine Learning, Digital campus, Academic performance, Augmented Education.