10th EAI International Conference on Simulation Tools and Techniques

Research Article

An Improved Similarity Calculation Algorithm Used in News Recommender System

  • @INPROCEEDINGS{10.1145/3173519.3173532,
        author={Zongwei Luo and Xinchen Shi},
        title={An Improved Similarity Calculation Algorithm Used in News Recommender System},
        proceedings={10th EAI International Conference on Simulation Tools and Techniques},
        publisher={ACM},
        proceedings_a={SIMUTOOLS},
        year={2018},
        month={8},
        keywords={recommender system; similarity calculation; jaccard algorithm; k-means algorithm},
        doi={10.1145/3173519.3173532}
    }
    
  • Zongwei Luo
    Xinchen Shi
    Year: 2018
    An Improved Similarity Calculation Algorithm Used in News Recommender System
    SIMUTOOLS
    ACM
    DOI: 10.1145/3173519.3173532
Zongwei Luo1,*, Xinchen Shi2
  • 1: Southern University of Science and Technology, Department of Computer Science and Engineering, Shenzhen, China
  • 2: Southern University of Science and Technology, Shenzhen, China
*Contact email: luozw@sustc.edu.cn

Abstract

Recommender system applies machine learning and focuses on solving the information explosion problem. The mainstream recommendation algorithms are: content based algorithm, collaborative filtering, and hybrid recommender system. All of them are facing the same problem, it is called similarity computation. Recommender systems usually use similarity to measure the items and users. In this paper, we first introduced some basic knowledge and the development trend of the recommender systems. Then we illustrate an improved similarity algorithm in news recommender system to make sure the algorithm pay more attention to users’ interest. The algorithm will modify their weight through users’ action. In the experiment, we use the AJS AWC-BC and AK-Means algorithms in real news recommender system and compared with some other algorithm. The result showed that the improved algorithm have better performance in vertical news recommend system.