Enhancing Adaptive Learning Systems: A Probabilistic Approach Using Bayesian Networks and User Clustering

Subject Areas : Multimedia Processing, Communications Systems, Intelligent Systems

Parvaneh Ahmadifar ¹ , Mojtaba Salehi ^{2 *} , Mohammad Yarahmadi ³ , Morteza Zakeri ⁴

1 - MSc, Department of Computer Engineering, Khorramabad Branch, Islamic Azad University, Khorramabad, Iran
2 - Instructor, Department of Computer Engineering, Khorramabad Branch, Islamic Azad University, Khorramabad, Iran
3 - Instructor, Department of Math, Khorramabad Branch, Islamic Azad University, Khorramabad, Iran
4 - Assistant Professor, School of Computer Engineering, Amirkabir University of Technology (Tehran Polytechnic), Tehran, Iran

Received: 2024-04-28 Accepted : 2024-06-09 Published : 2024-08-18

Keywords: Intelligent tutoring system, feature selection, Bayesian network, K-means algorithm,

Abstract :

Introduction: This study proposes an innovative framework to enhance the quality and adaptability of web-based educational systems using Bayesian networks. The proposed methodology aims to address issues like the cold-start problem and improve the scalability and personalization of educational services.

Method: The proposed methodology encompasses feature selection through the Relief algorithm, user clustering via k-means, and probabilistic service recommendation based on Bayesian inference. Feature selection prioritizes user-specific attributes, reducing data dimensionality and enhancing computational efficiency. Clustering organizes users into homogeneous groups based on similarities, facilitating tailored service provision. The system was validated using a dataset of 2000 users, each characterized by 16 features and linked to 24 services.

Results: The experimental results demonstrated an accuracy rate of 88.15%, underscoring the model's effectiveness compared to traditional techniques. Bayesian networks enabled the analysis of probabilistic relationships and uncertainties, ensuring that the most relevant services were recommended to each user.

Discussion: The proposed approach addressed the cold-start problem, improving scalability and adaptability to diverse user needs. Key advantages of this framework include enhanced personalization, computational efficiency, and significant improvements in user satisfaction. The study's findings highlight the potential of intelligent learning systems to adjust dynamically to user requirements, paving the way for more engaging and effective educational experiences. Future research will incorporate heuristic algorithms to improve prediction accuracy and integrate real-time user feedback to refine the recommendation process further.