Logo
ScienceStack
Table of Contents
Fetching ...
Sign In

Privacy-Preserving Personalization in Education: A Federated Recommender System for Student Performance Prediction

Rodrigo Tertulino, Ricardo Almeida

TL;DR

Education increasingly requires personalized content while respecting student privacy. The authors implement a privacy-preserving federated recommender system (RecommenderNet) trained with FedProx on the ASSISTments data, and compare it to a centralized XGBoost baseline. They show FedProx achieves a peak F1-score of 76.28%, about 92% of the centralized model's 82.85% F1, illustrating a favorable trade-off between privacy and performance. The work demonstrates practical potential for privacy-preserving adaptive learning platforms and outlines future directions including personalized federated learning and richer data modalities.

Abstract

The increasing digitalization of education presents unprecedented opportunities for data-driven personalization, but it also introduces significant challenges to student data privacy. Conventional recommender systems rely on centralized data, a paradigm often incompatible with modern data protection regulations. A novel privacy-preserving recommender system is proposed and evaluated to address this critical issue using Federated Learning (FL). The approach utilizes a Deep Neural Network (DNN) with rich, engineered features from the large-scale ASSISTments educational dataset. A rigorous comparative analysis of federated aggregation strategies was conducted, identifying FedProx as a significantly more stable and effective method for handling heterogeneous student data than the standard FedAvg baseline. The optimized federated model achieves a high-performance F1-Score of 76.28%, corresponding to 92% of the performance of a powerful, centralized XGBoost model. These findings validate that a federated approach can provide highly effective content recommendations without centralizing sensitive student data. Consequently, our work presents a viable and robust solution to the personalization-privacy dilemma in modern educational platforms.

Privacy-Preserving Personalization in Education: A Federated Recommender System for Student Performance Prediction

TL;DR

Education increasingly requires personalized content while respecting student privacy. The authors implement a privacy-preserving federated recommender system (RecommenderNet) trained with FedProx on the ASSISTments data, and compare it to a centralized XGBoost baseline. They show FedProx achieves a peak F1-score of 76.28%, about 92% of the centralized model's 82.85% F1, illustrating a favorable trade-off between privacy and performance. The work demonstrates practical potential for privacy-preserving adaptive learning platforms and outlines future directions including personalized federated learning and richer data modalities.

Abstract

The increasing digitalization of education presents unprecedented opportunities for data-driven personalization, but it also introduces significant challenges to student data privacy. Conventional recommender systems rely on centralized data, a paradigm often incompatible with modern data protection regulations. A novel privacy-preserving recommender system is proposed and evaluated to address this critical issue using Federated Learning (FL). The approach utilizes a Deep Neural Network (DNN) with rich, engineered features from the large-scale ASSISTments educational dataset. A rigorous comparative analysis of federated aggregation strategies was conducted, identifying FedProx as a significantly more stable and effective method for handling heterogeneous student data than the standard FedAvg baseline. The optimized federated model achieves a high-performance F1-Score of 76.28%, corresponding to 92% of the performance of a powerful, centralized XGBoost model. These findings validate that a federated approach can provide highly effective content recommendations without centralizing sensitive student data. Consequently, our work presents a viable and robust solution to the personalization-privacy dilemma in modern educational platforms.

Paper Structure

This paper contains 26 sections, 5 equations, 6 figures, 6 tables, 2 algorithms.

Table of Contents

  1. Introduction
  2. Related Work
  3. Theoretical Foundations
  4. Centralized Machine Learning
  5. eXtreme Gradient Boosting (XGBoost)
  6. Federated Learning
  7. Mitigating Statistical Heterogeneity with FedProx
  8. Methodology
  9. Dataset Description
  10. Data Preprocessing and Feature Engineering
  11. Experimental Setup and Models
  12. Centralized Model (Benchmark)
  13. Practical Application Framework
  14. Evaluation Metrics
  15. Results
  16. ...and 11 more sections

Figures (6)

  • Figure 1: Workflow of the Centralized Machine Learning approach. Sensitive student interaction data is collected from all clients and aggregated on a central server for model training.
  • Figure 2: The FL workflow. The process is iterative: (1) The server sends the global model to clients. (2) Clients train the model on their local, private data. (3) Clients send only the anonymized model updates back. (4) The server aggregates the updates to improve the global model.
  • Figure 3: A diagram illustrating the workflow of the federated recommender system in a practical student support scenario. The system leverages local and global data to generate a personalized learning path
  • Figure 4: Performance metrics of the centralized XGBoost model over 100 boosting rounds
  • Figure 5: Global performance metrics of the federated DNN model (FedProx, $\mu=0.5$) over 100 communication rounds, showing the convergence of Accuracy, Precision, Recall, and F1-Score
  • ...and 1 more figures