Back to Projects
Research Project
Saint Louis University

AI-Powered Educational Gaming & Analytics Platform

Intelligent gamification system that leverages machine learning to analyze student engagement patterns and optimize educational outcomes through adaptive game mechanics. Developed as a research initiative to revolutionize how educators understand and enhance student learning experiences.

Python
Machine Learning
Educational AI
Data Analytics
Gamification
TensorFlow
scikit-learn
Pandas
34%

Engagement Increase

91.7%

ML Model Accuracy

850+

Students Analyzed

28%

Performance Boost

Core Capabilities

Adaptive Learning Engine

Machine learning algorithms continuously analyze student performance and adjust game difficulty, content delivery, and challenge levels in real-time to maintain optimal learning zone.

Engagement Analytics

Advanced behavioral analytics track click patterns, session duration, task completion rates, and interaction frequency to identify engagement drop-offs and intervention opportunities.

Personalized Game Mechanics

Dynamic reward systems, achievement structures, and progression paths tailored to individual learning styles using clustering algorithms and preference prediction models.

Machine Learning Architecture

1

Data Collection & Feature Engineering

Capture multimodal student interaction data including click streams, time-on-task, error rates, help requests, and game progression. Extract 47 behavioral features using rolling window statistics and session aggregations.

2

Engagement Prediction Model

Random Forest classifier with 150 estimators predicts student engagement levels (High/Medium/Low) using behavioral features. Achieves 91.7% accuracy with SMOTE balancing for minority classes.

3

Student Clustering & Segmentation

K-Means clustering (k=5) identifies distinct learner personas based on engagement patterns, achievement preferences, and learning pace. Segments include Explorers, Achievers, Socializers, Strugglers, and Disengaged.

4

Adaptive Recommendation System

Collaborative filtering recommends optimal game mechanics, difficulty adjustments, and content sequences. Matrix factorization with SVD identifies latent factors predicting student-mechanic affinity.

5

Real-Time Intervention Triggers

Logistic regression model detects early warning signs of disengagement or frustration. Triggers adaptive interventions including difficulty reduction, hint delivery, or peer collaboration prompts.

Learner Personas Identified

Explorers

High curiosity, low achievement focus

24% of students • Prefer open-ended challenges and discovery-based mechanics

Achievers

Goal-oriented, high completion rates

31% of students • Respond to badges, leaderboards, and mastery challenges

Socializers

Collaboration-focused, peer interaction

18% of students • Thrive with team challenges and social comparison features

Strugglers

High effort, inconsistent progress

16% of students • Benefit from scaffolded hints and progressive difficulty

Disengaged

Low interaction, brief sessions

11% of students • Require immediate feedback and novelty-driven mechanics

Clustering Metrics

Silhouette Score: 0.68

K-Means with k=5 optimized via elbow method and silhouette analysis

Analytics Dashboard Features

Engagement Heatmaps

Visualize student activity patterns across time of day, day of week, and game modules using interactive heatmap grids. Identify optimal scheduling windows and content dead zones.

Learning Curve Analytics

Track individual and cohort-level mastery progression using exponential curve fitting. Predict time-to-competency and identify students falling below expected learning trajectories.

Intervention Impact Reports

A/B testing framework measures effectiveness of different game mechanics and intervention strategies. Statistical significance testing with Bonferroni correction for multiple comparisons.

Predictive Dropout Alerts

Logistic regression model with 87.3% precision identifies at-risk students 3-5 days before expected dropout. Auto-generates personalized re-engagement recommendations for educators.

Technology Stack

Machine Learning

  • • TensorFlow 2.13
  • • scikit-learn 1.3
  • • XGBoost
  • • SMOTE Balancing
  • • Surprise (SVD)
  • • SHAP Explainability

Data Processing

  • • Pandas
  • • NumPy
  • • Apache Spark
  • • PostgreSQL
  • • Redis Cache
  • • Airflow ETL

Visualization

  • • Plotly Dash
  • • Matplotlib
  • • Seaborn
  • • D3.js Integration
  • • React Charts
  • • Altair

Game Engine

  • • Phaser 3
  • • WebGL Rendering
  • • Socket.IO
  • • FastAPI Backend
  • • Docker Compose
  • • Nginx Proxy

Research Impact & Outcomes

Educational Effectiveness

34%

Average engagement increase across all student personas

28%

Improvement in assessment scores for adaptive group

41%

Reduction in student dropout rates during pilot semester

Research Contributions

  • Developed novel feature engineering approach for educational gaming analytics, achieving 91.7% accuracy in engagement prediction vs. 78% industry baseline
  • Identified five distinct learner personas using unsupervised clustering, providing actionable framework for personalized gamification strategies
  • Demonstrated statistically significant performance improvements (p < 0.001) across math, science, and language arts modules during 16-week pilot study
  • Created open-source analytics framework adopted by 3 educational institutions for research and classroom implementation

Future Research Directions

Deep Learning Integration

Exploring LSTM networks for temporal engagement pattern prediction and transformer models for content recommendation

Multimodal Learning Analytics

Integrating eye-tracking, facial expression analysis, and physiological signals for comprehensive engagement measurement