Inspiration I would like to detect students early who are likely to fail or withdraw as early as possible
What it does This is an early intervention engine. By analyzing the "Digital Heartbeat" of a student composed of seven data stream,It predicts the likelihood of failure or withdrawal with high accuracy before the first major assessment is even graded. It categorizes students into risk tiers, allowing educators to provide support exactly when it’s needed, rather than after the fact.
How we built it
We treated the OULAD structure as seven independent data silos, building a custom ETL (Extract, Transform, Load) pipeline to join them. Model Selection: We utilized a Multivariate Random Forest classifier. The decision was based on the model's ability to handle the non-linear relationship between a student’s background and their online behavior.
Challenges we ran into
The primary challenge was Temporal Alignment. Because we wanted an "Early Warning System," we couldn't use data from the end of the semester to predict the end of the semester. We had to "blind" our model at the 30-day and 60-day marks to ensure it could actually predict the future based only on a limited slice of the past.
Accomplishments that we're proud of
The 7 way Merge of various Successfully architecting a pipeline that flawlessly integrated seven disparate datasets into a single training matrix.
What we learned
We learned that Behavior Trumps History. While demographics (like socioeconomic status) provide context, the real-time interaction data—the frequency and duration of VLE clicks—was a much more powerful predictor of immediate risk
What's next for Student Perfomance Prediction System(SPPS)
The next phase for SPPS is Real-Time Integration. We want to move from a batch-processing model to a live API that can feed directly into a teacher's dashboard and work for various universities.
Log in or sign up for Devpost to join the conversation.