In order to improve the clustering accuracy of big data alliance users, this paper studies users' dynamic clustering based on their multi-dimensional attributes. First of all, the user profile of big data alliance is constructed from five dimensions of user basic attribute, domain attribute, preference attribute, social attribute and value attribute. And the K-means algorithm is used to cluster user profiles to complete the initial clustering. Then, based on the group user profile, combined with the user's recent dynamic behavior data, the FCM algorithm is used to achieve secondary clustering. Finally, the proposed user clustering method is tested by recommending data resources to the clustered user groups. The experimental results show that the user clustering method proposed in this paper has higher accuracy and lower error rate.
Wang, Xiaoxiao; Zhai, Lili; Hu, Yanling; and Zhang, Shuchen, "Research on Multi-Dimensional Dynamic Clustering Method of Big Data Alliance Users" (2020). WHICEB 2020 Proceedings. 29.