Use of similarity measure in recommender system based on type of item preferences

Abstract

During last twenty years recommender system have emerged as a research field. Recommender System is rooted in the field of Information Retrieval, Machine Learning and Decision Support System. Most of the users do not have enough knowledge to make automatic decisions. So they need recommendation of different items for better choice. Because of this many researchers tried to understand the algorithmic techniques for recommendation to the given user. It is very important factor to identify similar items related to the target user’s test. To find similar items RS uses item preference of an item. In different RSs, the item preferences are available in different forms, i.e. preferences are either available, Boolean preference (yes/no) or not available. We test various User Similarity Measures for dataset with preferences, without preferences and Boolean preferences. We tested various similarity measures for User Based Collaborative Filtering techniques in Apache Mahout.