An ensemble model approach for many-feature data clustering

Abstract

The ensemble is a popular machine learning technique based on the principle of divide and conquer. In data clustering, the ensemble aims to improve performance in terms of processing speed and clustering quality. Most existing ensemble methods face inherent complex challenges such as uncertainty, ambiguity, and overlap. Fuzzy clustering has recently been developed to handle data with many-feature, heterogeneity, uncertainty, and big data. In this paper, we propose an ensemble feature- reduction clustering model (EFRC) using advanced machine learning techniques. The EFRC model consists of three phases. First, the data is feature-reduced by a random projection. Then, the data is divided into subsets based on the likelihood of overlap and quantification of noise. Various clustering techniques are used to cluster the subset of data. Finally, the results of the clustering modules are consensus using the classification technique to produce the final clustering result. Several tests were performed on the benchmark datasets. The test results show the superior performance of the EFRC model compared to the previous models.