Student Dropout Risk Analysis
An analysis of 4,400+ student records across two real-world datasets that identified first-semester GPA as the dominant dropout predictor, and turned it into actionable institutional recommendations.
The Problem
Institutions collect mountains of academic, demographic, and socioeconomic data on students, but without analysis it can't answer the question that matters: who is about to drop out, and when can we still intervene?
What I Built
A full analysis over the UCI Dropout and OULAD datasets: IQR-based outlier detection across six continuous features (with a justified decision to retain genuine variation), correlation analysis, and multi-group boxplot comparison across academic and socioeconomic variables.
The Result
First-semester GPA emerged as the dominant predictor while admission grades, age, GDP, and unemployment proved weak, translated into concrete recommendations: early-semester risk flagging, targeted intervention triggers, and evidence-based resource allocation.