+44 1300 500008
Recommender Systems are software tools and techniques that seek to suggest items that are likely of interest to a particular user. These systems are a key part of most e-commerce applications, as they ease users to find products that meet their needs while improving sales. Several approaches have been created to determine the users’ preferences by working with different sources and types of information. Collaborative filtering uses the history of ratings, content and knowledge based recommenders work with the features of items, context aware systems provide suggestions based on the situation parameters or conditions that surround the user, while demographic filtering utilizes user’s demographic characteristics. Additionally, there are hybrid approaches that fuse two or more techniques in order to overcome the shortcomings of each method.
In this work an application of dynamic selection to the recommender systems field is studied. This selection strategy, taken from Multiple Classifier Systems, consists of selecting a specific set of classifiers for each test pattern. To adapt this notion to the context of this research, it was proposed a hybrid system that dynamically seeks to select the best recommendation method in each prediction.
After carrying out experiments, the application of dynamic selection did not provide any significant improvement to recommendations. However, the inclusion of demographic and contextual information in a hybrid content-based basis increased the accuracy of the system considerably. The final solution was evaluated using datasets containing reviews of hotels and books. Results showed that the recommender is capable of working in tourism related scenarios and that can also be parameterized to other recommendation problems as long as content, demographic or contextual features are available.