Authors:
Nabil Belacel
1
;
2
;
Guangze Wei
1
and
Yassine Bouslimani
1
Affiliations:
1
Department of Electrical Engineering, Moncton University, New Brunswick, Canada
;
2
Digital Technologies Research Centre, National Research Center, Ottawa, Ontario, Canada
Keyword(s):
Machine Learning, Supervised Learning, Classifier, Information Retrieval, Content Based Filtering, Recommender System.
Abstract:
This paper presents the application of classification method based on outranking approach to Content Based Filtering (CBF) recommendation system. CBF intends to recommend items similar to those a given user would have liked in the past by first extracting traditional content features such as keywords and then predicts user preferences. Therefore content based filtering system recommends an item to a user based upon a description of the item and a profile of the user’s interests. Typically, to represent user’s and items’ profiles the existing CBF recommendation systems use the vector space model with basic term frequency and inverse document frequency (tfidf ) weighting. The tfidf and cosine similarity techniques are able, in some cases, to obtain good performances, however, they do not handle imprecision of features’ scores and they allow the compensation between features which will lead to bad results. This paper introduces k Closest resemblance classifier for CBF. The detailed mode
ls in this paper were tested and compared with the well-known tfidf based the k Nearest Neighbor classifier using Amazon fine food and book reviews data-set. The preliminary results show that our proposed model can substantially improve personalized recommendation of items described with short text like products description and customers’ review.
(More)