Loading...
Please wait, while we are loading the content...
Similar Documents
Towards a reference architecture for context-aware recommender systems
| Content Provider | Semantic Scholar |
|---|---|
| Author | Keijers, Bm Bram |
| Copyright Year | 2014 |
| Abstract | Recommender systems have become increasingly important in web systems as a tool to overcome the information overload problem. Research has and will be an important factor in the development of recommender systems, focusing on opportunities to provide more personalized recommendations to the user. Incorporating contextual information from the users in the recommendation process has been one of the recent research directions, resulting in Context-Aware Recommender Systems (CARS). In the future acquisition, representation and evaluation of contextual information will remain an active research direction, as shown in recent literature [VMO+11]. Recommender systems are often specifically built towards a domain, or serving web system. These systems often adhere to the identified requirements and threats towards the recommendation quality. Unforeseen problems in these recommender systems can be hard to solve when the implementation is too specific. Moreover, one of the requirements of CARS is having a general strategy enabling the availability of context throughout the system. This is required in order to be able to optimal use the available contextual information in all the processing tasks of the CARS. In both cases flexibility and adaptability in the recommendation processes is required. In this thesis a Reference Architecture (RA) for CARS is described that provides a general system outline which adheres to these requirements. This is done by identifying the main components and providing separation between core processing and recommendation deriving tasks. A literature analysis has been conducted to obtain main references and components so that a state of the art RA could be constructed. In this analysis requirements and components of recommender systems are inventoried. Furthermore active research in the recommender system field is referenced in order to be able to adhere to these requirements as well. The literature analysis is supported by a system review of the Masters Portal web system from Study Portals. In this way practical information of an actual live web system is used as well. In the system review the opportunities of context application and CARS are inventoried, generalized and used in the RA. The evaluation of quality, correctness and effectiveness of the RA is done by implementing parts of the RA. The evaluation resulted in feedback on the completeness of the objects and components in the RA. Additionally, the implementations are used in practice in an exploratory data analysis with the Context-Aware Recommendation Adjustment (CARA) algorithm. This algorithm composes a new ranking of a recommendation based on usage behavior. Contextual information is used to provide separation in the usage data and to capture similar preferences of users having corresponding |
| File Format | PDF HTM / HTML |
| Alternate Webpage(s) | https://pure.tue.nl/ws/files/46957887/772075-1.pdf |
| Alternate Webpage(s) | http://alexandria.tue.nl/extra1/afstversl/wsk-i/keijers2014.pdf |
| Language | English |
| Access Restriction | Open |
| Content Type | Text |
| Resource Type | Article |
