A New Fuzzy Cluster Forests Method For Big Data

With the accumulation of the large data size, clustering of big data is a challenging task. However, data reduction is considered as a powerful model which significantly reduces execution time. This work presents a new fuzzy clustering ensemble model based on cluster forests method (CF) that can simultaneously reduce the execution time and consists mainly of two steps: generation of clusters instances and aggregation of global models. In the beginning, this algorithm makes multiple clusters instances using fuzzy clustering bdrFCM technique. Secondly, it aggregates this clusters to obtain final results using Ncut spectral clustering. We call it as FCE-CF approach. This proposed method is guided by cluster validity index kappa. Experimental results demonstrate that the FCE-CF outperforms the existing clustering methods in terms of time and memory on big data UCI repository.