A framework for generating recommendations based on trust in an informal e-learning environment.

Area of Science:

• Computer Science
• Information Science
• Educational Technology

Background:

• Online learning environments face information overload and data sharing challenges.
• The learner cold-start problem hinders effective question answering for new users in e-learning communities.
• Existing recommendation systems struggle with sparsity and accuracy in informal learning contexts.

Purpose of the Study:

• To propose a novel Trust-aware Deep Neural Recommendation (TDNR) framework to address learner cold-start issues.
• To enhance recommendation accuracy by modeling complex nonlinear relationships and incorporating trust degrees.
• To facilitate the formation of informal e-learning communities through improved expert-questioner matching.

Main Methods:

• Developed a Trust-aware Deep Neural Recommendation (TDNR) framework.
• Utilized Latent Dirichlet Allocation (LDA) for tag modeling and expert ranking using hub and authority scores.
• Incorporated a questioner-responder relational graph and trust degrees into the recommendation model.
• Evaluated the model on the Stack Overflow dataset using MAP, RMSE, and F-measure metrics.

Main Results:

• The TDNR framework significantly improves recommendation accuracy compared to rating-based and social-trust-based approaches.
• TDNR demonstrates superior performance in Mean Absolute Error (MAE), Root Mean Squared Error (RMSE), and F-measure metrics.
• The model effectively addresses information overload and user sparsity, providing personalized recommendations in cold-start scenarios.

Conclusions:

• The proposed TDNR framework offers a robust and reliable solution for learner cold-start problems in e-learning.
• TDNR enhances user experience by facilitating supportive learning communities and accurate information retrieval.
• The hybrid approach effectively models complex relationships, leading to more relevant and personalized recommendations.