Résumé

Internet users are overwhelmed by the vast number of services and products to choose from. This data deluge has led to the need for recommender systems. Simultaneously, the explosion of interactions on social networks is constantly increasing. These interactions produce a large amount of content that incites organizations and individuals to exploit it as a support for decision making. In our research, we propose, Twit-CoFiD, a hybrid recommender system based on tweet sentiment analysis which performs a demographic filtering to use its outputs in an enhanced collaborative filtering enriched with a sentiment analysis component. The demographic filtering, based on a Deep Neural Network (DNN), allows to overcome the cold start problem. The sentiment analysis of Twitter data combined with the enhanced collaborative filtering makes recommendations more relevant and personalized. Experiments were conducted on 1M and 100K Movielens datasets. Our system was compared to other existing methods in terms of predictive accuracy, assessed using Root Mean Square Error (RMSE) and Mean Absolute Error (MAE) metrics. It yielded improved results, achieving lower RMSE and MAE rates of 0.4474 and 0.3186 on 100K Movielens dataset and of 0.3609 and 0.3315 on 1M Movielens dataset.

Résumé

The rapid expansion of contextual information from smartphones and Internet of Things (IoT) devices paved the way for Context-Aware Recommendation Systems (CARS). This abundance of contextual data heralds a transformative era for traditional recommendation systems. In alignment with this trend, we propose a novel model which provides personalized recommendations based on context. Our approach uses K-means algorithm to cluster users based on contextual features. Then, the model performs collaborative filtering based on matrix factorization with enhanced contextual biases to provide relevant recommendations. We demonstrated the performance of our method through experiments conducted on the movie recommender dataset LDOS-CoMoDa. The experimental results showed the effective performance of our proposal compared to reference methods, achieving an RMSE of 0.7416 and an MAE of 0.6183.

Résumé

The cold start problem has always been a major challenge for recommender systems. It arises when the system lacks rating records for new users or items. Addressing the challenge of providing personalized recommendations in the cold start scenario is crucial. This research proposes a new hybrid recommender system named CoDFi-DL which combines demographic and enhanced collaborative filtering. The demographic filtering is performed through a deep neural network (DNN) and used to solve the new user cold start problem. The enhanced collaborative filtering component of our model focuses on delivering personalized recommendations through a neighborhood-based method. The major contribution in this research is the DNN-based demographic filtering which overcomes the new user cold start problem and enhances the collaborative filtering process. Our system significantly improves the relevancy of the recommendation task and thus provides personalized recommended items to cold users. To evaluate the effectiveness of our approach, we conducted experiments on real multi-label datasets, 1M and 100K MovieLens. CoDFi-DL recommender system showed higher performance in comparison with baseline methods, achieving lower RMSE rates of 0.5710 on the 1M MovieLens dataset and 0.6127 on the 100K MovieLens dataset.

Résumé

In a rapidly evolving digital landscape, recommender systems have become essential tools for helping users navigate overwhelming amounts of information in various domains. In e-learning contexts, these systems aim to support learners by identifying educational resources or relevant academic content that are significant for their learning experience. However, research incorporating domain knowledge, such as course-related concepts, to improve recommendation quality remains limited. This paper presents a new personalized recommender system in e-learning context, which is called KA-ERN: Knowledge-based Attention Ensemble Recurrent Network. Specifically, KA-ERN leverages a Knowledge Graph to capture the dependencies and semantic relationships between users, courses, and their related concepts and then, performs ensemble learning by combining the Bidirectional Long Short-Term Memory network (Bi-LSTM) with the Artificial Neural Network (ANN). An attention mechanism is added to enhance recommendation quality. Our approach is based on a two-stage architecture. First, entities and relations embeddings are generated and then concatenated as sequences allowing the model to capture complex relationships and contextual dependencies of user preferences. Secondly, these embeddings are provided as inputs to the KA-ERN model. The proposed combination of Bi-LSTM network, ANN, and attention mechanisms shows the advantage of using Knowledge graph embeddings over bipartite graph embeddings capturing only user-item interactions and experimental results achieving the best recommendation metrics, with RMSE of 0.045 and MAE of 0.034, outperforming baseline methods.

Résumé