An Approach to Increase Efficiency of Opinion Mining and Recommendation Collaborative Filtering System using Similarity-based Algorithms

Increasingly increasing social networking has led many users to take that part and often manage their daily activities on these networks. The traditional similarity criteria, such as the Pearson correlation coefficient (PCC, cosine similarity (COS), and mean square difference (MSD), assume similarity of a symmetric concept, that is, two users have the same effect on each other.) A subtle point that is worrying is ignoring the criteria the similarities between the different parts of the items are shared between the two users. Equivalence functions produce the equivalence ratio for two users because they penalize both users for different parts of the shared items. Recommender systems are one of the most successful web-based personalization tools. The most important task of an advisory system is to discover the user s favorite items. Extremely large spaces are selectable items. Similarity algorithms are often considered as collaborative refinement techniques based on memory and one of the most successful methods in advisory systems. When explicit the similarity is usually defined using similarity functions, such as Pearson correlation coefficient, cosine similarity, or mean square difference. These criteria assume that the similarity is symmetric, so that the two users introduce new equivalence effects on Have one another in this thesis, we introduce new weighing schemes that allow us to consider new features to find similarities between users. These weigh-in schemes first convert the symmetric similarity to asymmetric similarity by considering the number of users votes to the items. Then, the impact of user habits on ranking items is considered by measuring the number of repetitions of each rank for the ranked items in common. Experiments will be performed on the data set and the results will be compared with similarity criteria. The results will show that the addition of weightedmethods to the traditional similarity criteria significantly improves the results of them.

Keywords:

Advisor system , Opinion mining system , Participatory filtering , Social networking , Similarity algorithms.