← Back to projects

Predictive Analysis of Student Dropout Risk in an EdTech Platform

Using engagement and performance data to identify at-risk learners early and improve course completion.

Overview

This project applies data science to improve learning outcomes in an EdTech environment. Using behavioral and engagement data, I developed a predictive model to identify students most likely to drop out before course completion.

Objectives

  • Predict which students are at risk of non-completion early in the learning journey
  • Identify engagement metrics most correlated with success
  • Recommend data-driven interventions to improve retention and learning outcomes

Approach

Data Preparation & Cleaning

  • Gathered and merged datasets containing student demographics, activity logs, quiz performance, and course progress.
  • Cleaned missing data, encoded categorical variables, and engineered features (avg. session duration, quiz streaks, days inactive).

Exploratory Data Analysis (EDA)

  • Visualized learning behavior trends using Seaborn and Matplotlib.
  • Found key patterns — e.g., inactivity > 5 days correlated with a 70% chance of dropping out.

Model Development

  • Built Logistic Regression and Random Forest models using Scikit-learn.
  • Achieved 86% accuracy in predicting potential dropouts.
  • Identified top predictors: activity frequency, quiz score variance, and time spent per module.

Insights & Recommendations

  • Recommend early engagement nudges for low-activity users.
  • Provide personalized quiz feedback to improve confidence and retention.
  • Integrate dashboards for educators to monitor and support at-risk students.

Tech Stack

Python (Pandas, NumPy, Scikit-learn, Matplotlib, Seaborn) · SQL · Excel

Impact

Helped design proactive interventions that improved student completion rates by 14%, strengthening learner success and satisfaction.