Fast mining method of network heterogeneous fault tolerant data based on K-means clustering

Article type: Research Article

Authors: Huang, Haiyang^{; *} | Shang, Zhanlei

Affiliations: Engineering Training Center, Zhengzhou University of Light Industry, Zhengzhou 450001, China

Correspondence: [*] Corresponding author. E-mail: yonghuangk@aliyun.com.

Abstract: In the traditional network heterogeneous fault-tolerant data mining process, there are some problems such as low accuracy and slow speed. This paper proposes a fast mining method based on K-means clustering for network heterogeneous fault-tolerant data. The confidence space of heterogeneous fault-tolerant data is determined, and the range of motion of fault-tolerant data is obtained; Singular value decomposition (SVD) method is used to construct the classified data model to obtain the characteristics of heterogeneous fault-tolerant data; The redundant data in fault-tolerant data is deleted by unsupervised feature selection algorithm, and the square sum and Euclidean distance of fault-tolerant data clustering center are determined by K-means algorithm. The discrete data clustering space is constructed, and the objective optimal function of network heterogeneous fault-tolerant data clustering is obtained, Realize fault-tolerant data fast mining. The results show that the mining accuracy of the proposed method can reach 97%.

Keywords: K-means clustering, network heterogeneous fault-tolerant data, fast mining, redundancy, unsupervised feature selection algorithm

DOI: 10.3233/WEB-210460

Journal: Web Intelligence, vol. 19, no. 1-2, pp. 115-124, 2021