EARLI Project

Introduction 

The Early AleRt for Learning Intervention (EARLI) project by ATLAS leverages four machine and deep learning approaches to predict student in need of academic support before the new semester. In essence, four different models (SVM, logistic regression, LSTM and Short-term gated LSTM) were built based on historical performance data. The combined classification of every incoming second- and third-year student from these models is used to identify the students most at risk and the course(s) where attention is/are needed. 

Impact 

About twenty to forty most at-risk students are identified by the algorithm for each school. This list of students is then given to student care managers via a dashboard, who would then reach out to the students. Currently five STEM schools (CEE, EEE, MAE, MSE and SPMS) were involved in the project. The true positive rate is more than 70% from surveyed schools.  

Here is a testimony from CEE’s Associate Chair (Students): 

"The EARLI Dashboard has proven invaluable in pre-emptively identifying students at risk of struggling in specific courses before enrolment. Prior to its implementation, pinpointing students facing potential challenges was difficult, given the vast student body and extensive range of courses available. With this tool, greater attention can be directed towards the identified students to provide them with the necessary support and assistance to succeed."

In AY23/24 semester, MAE reported that 13 out of 17 “true positive” students who were reached out by the school care manager saw an increase in their semester GPA and cumulative GPA. One such student went from failing 4 of his core modules to obtaining an average of B’s for the semester after the school’s intervention. Another student felt that the support given by the school really gave him hope to continue pushing on in his studies and graduate. He went from a sGPA of 1.38 to 2.82, from mostly D grades to B’s and C’s.

Below is his feedback:

“I feel that it is very helpful as it gives students like me confidence and hope to make improvements in my studies knowing the school is supporting and watching over us. Before the call I thought the school simply do not care about students that are performing badly in their studies. The call definitely gave me more drive to achieve my goal of pulling my GPA up to at least 2.5 at graduation.” - MAE Student 

#PredictiveAI #at-risk student prediction #aiforeducation