IKM-NCS: A Novel Clustering Scheme Based on Improved K-Means Algorithm
Journal: Engineering World (Vol.1, No. -)Publication Date: 2019-12-31
Authors : Weipeng Wang Shanshan TU; Xinyi Huang;
Page : 103-108
Keywords : clustering algorithms; k-means; feature extraction; behavior analysis;
Abstract
—Aiming at the problems of distorted center selection and slow iteration convergence in traditional clustering analysis algorithm, a novel clustering scheme based on improved k-means
algorithm is proposed. In this paper, based on the analysis of all user behavior sets contained in the initial sample, a weight calculation method for abnormal behaviors and an eigenvalue extraction method for abnormal behavior set are proposed and a set of abnormal behaviors is constructed for each user according to the behavior data generated by abnormal users. Then, on the basis of the traditional k-means clustering algorithm, an improved algorithm is proposed. By calculating the compactness of all data points and selecting the initial cluster center among the data points with high and low compactness, the clustering performance is enhanced. Finally, the eigenvalues of the abnormal behavior set are used as the input of the algorithm to output the clustering results of the abnormal behavior. Experimental results show that the clustering performance of this algorithm is better than the traditional clustering algorithm, and can effectively improve the clustering performance of abnormal behavior.
Other Latest Articles
- A Compact Gradient Based Neural Network for Capon Spectral Estimation
- Tetrolet Local Directional Pattern and Optimization-driven 2D-HMM for Face Recognition
- Systematic and robust air cleanser for cleaning a pollution caused by the Rocket Stove
- Analytical approach to a three species food chain model by applying Homotopy perturbation method
- Empirical investigation of noise reduction filter for a flow-based spirometer accuracy improvement
Last modified: 2020-08-04 23:23:36