Learning Similarity from Collaborative Filters
نویسندگان
چکیده
Collaborative filtering methods (CF) exploit the wisdom of crowds to capture deeply structured similarities in musical objects, such as songs, artists or albums. When CF is available, it frequently outperforms content-based methods in recommendation tasks. However, songs in the so-called “long tail” cannot reap the benefits of collaborative filtering, and practitioners must rely on content-based methods. We propose a method for improving contentbased recommendation in the long tail by learning an optimized similarity function from a sample of collaborative filtering data. Our experimental results demonstrate substantial improvements in accuracy by learning optimal similarity functions.
منابع مشابه
A New Similarity Measure Based on Item Proximity and Closeness for Collaborative Filtering Recommendation
Recommender systems utilize information retrieval and machine learning techniques for filtering information and can predict whether a user would like an unseen item. User similarity measurement plays an important role in collaborative filtering based recommender systems. In order to improve accuracy of traditional user based collaborative filtering techniques under new user cold-start problem a...
متن کاملA NOVEL FUZZY-BASED SIMILARITY MEASURE FOR COLLABORATIVE FILTERING TO ALLEVIATE THE SPARSITY PROBLEM
Memory-based collaborative filtering is the most popular approach to build recommender systems. Despite its success in many applications, it still suffers from several major limitations, including data sparsity. Sparse data affect the quality of the user similarity measurement and consequently the quality of the recommender system. In this paper, we propose a novel user similarity measure based...
متن کاملDesigning collaborative learning model in online learning environments
Introduction: Most online learning environments are challenging for the design of collaborative learning activities to achieve high-level learning skills. Therefore, the purpose of this study was to design and validate a model for collaborative learning in online learning environments. Methods: The research method used in this study was a mixed method, including qualitative content analysis and...
متن کاملThe Collaborative Learning in the e-Learning Environments
Introduction: The collaborative learning and interactive electronic-learning (e- learning) is one of the key factors in education system success. This study examined the collaborative e-learning in the framework of constructivism theory. Methods: This is a review article. The databases such as Scientific Information Databases, Magiran, Science Direct, and Google Scholar were reviewed. Also,...
متن کاملCollaborative Datasets Retrieval for Interlinking on Web of Data
Dataset interlinking is a great important problem in Linked Data. We consider this problem from the perspective of information retrieval in this paper, thus propose a learning to rank based framework, which combines various similarity measures to retrieve the relevant datasets for a given dataset. Specifically, inspired by the idea of collaborative filtering, an effective similarity measure cal...
متن کامل