Author : Naji Nawaf, Huda
journal of kerbala university,
2016, Volume 12, Issue 1, Pages 254-260
This study intends to improve the recommendation system by unifying both the implicit and explicit behavior of users. The implicit behavior indicates what users view over time regardless of the rating of what he views (implicit rating), whereas, the explicit behavior in this effort refers to what users rate for an item. The hamming distance is used to create a distance matrix that is converted to similarity matrix for users who are akin in terms of implicit rating. As for explicit ratings, the cosine similarity is used to create similarity matrix for users who are similar in terms of the scale used to rate an item. The proposed method is evaluated using three data sets; Movielens, Hetrec 2011, and Yahoo! Movies. The evaluation of the proposed method constrains with the measures of the related works. Thus, recall, precision, mean absolute error (MAE), and F-measures have been used. The experimental results show that the proposed system has a good performance, particularly Movielens dataset, when compared with the existing works. Our proposed method is free of any complex computations; at the same time it is competitive as comparable and better results are obtained considering other works.
Categories and Subject Descriptors: [Computer-Social Networks]