An Implementation of High Performance Parallel KNN Algorithm Based on GPU

Authors:
  • LHLican Huang
    Lican Huang
  • ZLZhen Liu
    Zhen Liu
  • ZYZhiwen Yan
    Zhiwen Yan
  • PLPiaoyue Liu
    Piaoyue Liu
  • QCQingpeng Cai
    Qingpeng Cai

Abstract
Summary form only given, as follows. Tremendous data will be generated in ERP system of an enterprise. Those data are very valuable for the enterprise. Therefore, Business Intelligent (BI) is used to help companies. Business Intelligent often uses data mining technologies. One of the methods in data mining is k-nearest neighbor (KNN). The time complexity of KNN is a bottleneck in the practical application in the fields of intelligent business. GPU has obvious advantages for enhancing effectiveness by parallelizing the KNN algorithm. There are lots of researches about GPU based KNN such as the work of V. Garcia, et.al.. However, we focus the GPU based KNN computation on the domain of Business Intelligent.