Computer and Modernization

Previous Articles     Next Articles

Collaborative Filtering Recommendation Algorithm #br# Based on Multifactor Expert Group Scoring

  

  1. Computer Center, Department of Experiment Teaching, Guangdong University of Petrochemical Technology, Maoming 525000, China
  • Received:2015-04-14 Online:2015-07-23 Published:2015-07-28

Abstract:

In view of the limitations of traditional collaborative filtering recommendation algorithm, this paper puts forward a new collaborative filtering recommendation
algorithm. Firstly, the paper uses the global professional index and local active index to define the requirement of experts; secondly, selects appropriate proportions users to
constitute experts group, and then assigns weights for each expert according to the expert’s judgment and dissimilarities between experts and target users, finally, defines the
predictive scores. Meanwhile, members of the group are dynamic, experts have different weights, so the recommended results are closer to the target user. Because the information
utilization is high, it can get a clear result. The experimental results on open dataset named GroupLens and Netflix show that, the algorithm in prediction success rates is
superior to the traditional method.

Key words: global professional index, local active index, dissimilarities, judgment, collaborative filtering recommendation algorithm

CLC Number: