Ph.D Thesis description
Recommender systems have been used in many domains to assist users' decision making by providing item recommendations and thereby reducing information overload. Context aware recommender systems (CARS) go further, taking contexts (e.g., time, location, occasion, etc) into consideration to suggest items that are appropriate in users specific contextual situations.
We aim to develop a Context-Aware recommendation system. Our key contributions are as follows:
- A weighting method that takes into account the different dimensions and their correlations.
- A context-aware situation prediction based on relevant dimensions.
- Recommendation according to user situation.
The prediction task has been highlighted in the literature as the most important one computed by context-aware recommendation algorithms, where the goal is to predict the rating that a user would assign to an item that he/she has not rated according to his/her current situation. Although rating prediction based recommender systems are widely popular, two thriving challenges are still to tackle: (i) context dimension s weighting and (ii) correlation between context dimensions. Several contextual recommendation algorithms have been developed by incorporating context into recommenders in different ways. Most of those recommendation algorithms ignore the weight of importance of each context dimension and correlations that may exist among them. A number of contextual dimensions have been identified as important in different recommendation applications: such as companion in the movie domain, time and mood in the music domain, and weather or season in the travel domain. To address these issues, we propose a novel approach for weighting individual context dimension and combination of dimensions, studying the correlation between them to infer the current situation then predict the rating according to the situation in which the user is involved. In order to evaluate our method, we have created two larges contextual datasets, due to the fact that the most publicly available context-aware datasets are small and sparse.