An Evaluation of Neighbourhood Formation on the Performance of Collaborative Filtering

摘要: Personalisation features are key to the success of many web applications and collaborative recommender systems have been widely implemented. These assist users in finding relevant information or products from vast quantities that frequently available. In previous work, we demonstrated such vulnerable attack recommendations can be manipulated. We introduced concept robustness as a performance measure, which is defined ability system provide consistent predictions presence noise data. this paper, expand on our work by examining effects several neighbourhood formation schemes similarity measures performance. propose filtering mechanism for false profiles order improve system.