Πανεπιστήμιο Μακεδονίας
Recommendations in mobile commerce environments: supporting quality and privacy requirements
Abstract
dc:descriptionRecommender systems have become a technology that has been widely used by various online applications in situations where there is an information overload problem. Numerous applications such as e-commerce, movie recommendations, song recommendations and people to people recommendations in social networks helping the users, which has improved the quality of the experience and revenues of the vendor. The development of recommender systems has been focused mostly on the development of algorithms, such as collaborative filtering, that provide more accurate recommendations in web environments. However, the use of mobile devices and the rapid growth of the internet and networking infrastructure has brought the necessity of using mobile recommender systems in different environments and scenarios. In mobile environments, every recommendation system is developed with a specific task in mind since a mobile recommendation domain does not exist now. Furthermore, when a mobile recommender system is developed a set of components derived from traditional web recommender systems is used. However, little work has been done towards the identification of the requirements that have an impact on the development of mobile recommender systems.In this thesis, the factors that affect the quality of recommender systems in mobile environments have been extensively analyzed by the author and novel work towards mobile recommender systems development is presented. To that end, the author assists in the development of mobile recommender systems by analyzing the factors that affect the recommendations in mobile scenarios and then makes four contributions: In the first contribution, the concept of static multi-level collaborative filtering for the improvement of the recommendations is introduced. In the second contribution, a dynamic multi-level collaborative filtering is proposed. In the third contribution, a method for privacy-preserving collaborative filtering recommendations is proposed, while in the fourth contribution a method for privacy-preserving context-aware mobile recommendations is presented. The proposals have been tested on common experimental designs that consider real datasets from different recommendation domains, widely used evaluation metrics and comparisons to alternative methods where necessary. The results of the experiments support the validity of the contributions and provide useful insights on their behavior, while supporting their practical applicability.
Degree
thesis:*- Grantor dc:publisher
- Πανεπιστήμιο Μακεδονίας
- Year dc:date
- 2017
Author and committee
dc:creator, dc:contributor.*- Authors dc:creator
-
- Polatidis, Nikolaos
- Πολατίδης, Νικόλαος
Subjects
dc:subject × 12Rights
dc:rights- Statement dc:rights
-
- BY
- Language dc:language
- eng
Identifiers
dc:identifier.*- Identifier
- 10.12681/eadd/44490