هیلدا احمدی - موسسه آموزش عالی لیان بوشهر
مرضیه دادور - موسسه آموزش عالی لیان بوشهر
حسن ارفعی نیا - موسسه آموزش عالی لیان بوشهر

Data plays an important role in everyday life. The existence of the Internet has caused a significant growth of data and the creation of big data. Managing and analyzing this data, which is often unlabeled, is a big challenge for the real world. Hierarchical clustering is an unsupervised learning algorithm for grouping data points with similar characteristics. In this article, a cumulative hierarchical clustering algorithm based on clustering of clusters and map-reduce model is introduced. The main idea of cumulative clustering is proposed by combining the results of different individual clustering methods. Cumulative methods lead to higher quality in creating clusters due to multiple learning. In addition, the reduce mapping model for clustering aggregation is a model for clustering big data. The proposed algorithm consists of three main steps. Selection of a subset of single hierarchical clustering methods and the data are clustered by selected single hierarchical clustering methods. In the second step, the created clusters are re-clustered to create hyper-clusters. After clustering the clusters, each sample is assigned to a hyper-cluster with maximum similarity. Based on this, the final clusters are formed in the last stage. Simulation has been done on several datasets from the UCI repository and the results show the better performance of the proposed method compared to methods such as CHC and RCESCC.

Hierarchical clustering, AHC, similarity measure, consensus function, clustering of clusters