A Fast Similarity Algorithm for Personal Ontologies Using Triangle Inequality
碩士 === 國立中興大學 === 資訊科學與工程學系所 === 97 === The Personal Ontology Recommender System (PORE) currently operated in the library of National Chung Hsing University is a recommender system developed by our research team. The system consists of content-based recommendation model based on personal ontology an...
Main Authors: | , |
---|---|
Other Authors: | |
Format: | Others |
Language: | zh-TW |
Published: |
2009
|
Online Access: | http://ndltd.ncl.edu.tw/handle/93888608461842522310 |
Summary: | 碩士 === 國立中興大學 === 資訊科學與工程學系所 === 97 === The Personal Ontology Recommender System (PORE) currently operated in the library of National Chung Hsing University is a recommender system developed by our research team. The system consists of content-based recommendation model based on personal ontology and collaborative filtering recommendation model. For collaborative filtering, the recommender system needs to compute the similarity between any two users. That will incur lots of computations because the library currently has more than thirty thousands of users and three hundred thousands of collections.
The purpose of this thesis is to design an efficient algorithm for computing the similarity between two users. A personal ontology representing the favorites of a user in PORE is a tree structure. In this thesis, we define tree distance for measuring the dissimilarity between two users. We then propose an efficient algorithm for calculating ontology similarities using triangle inequality. The experimental results show that the proposed method can save up to 88% of comparisons compared to that of brute force algorithm.
|
---|