An Entropy-Based Similarity Measure for Collaborative Filtering
Autor: | Soojung Lee |
---|---|
Rok vydání: | 2018 |
Předmět: |
business.industry
Computer science 02 engineering and technology Similarity measure Recommender system Machine learning computer.software_genre Global Rating 020204 information systems 0202 electrical engineering electronic engineering information engineering Collaborative filtering Entropy (information theory) 020201 artificial intelligence & image processing Artificial intelligence business computer |
Zdroj: | Data Mining and Big Data ISBN: 9783319938028 DMBD |
Popis: | Collaborative filtering is a successfully utilized technique in many online commercial recommender systems. Similarity measures play an important role in this technique, as items preferred by similar users are to be recommended. Although various similarity measures have been developed, they usually treat each pair of user ratings separately or simply combine additional heuristic information with traditional similarity measures. This paper addresses this problem and suggests a new similarity measure which interprets user ratings in view of the global rating behavior on items by exploiting information entropy. Performance of the proposed measure is investigated through various experiments to find that it outperforms the existing similarity measures especially in a small-scaled sparse dataset. |
Databáze: | OpenAIRE |
Externí odkaz: |