An Evolutionary Stream Clustering Technique Outlier Detection in Medical Data

Abstract

AbstractClustering data streams from data mining become evident as one of the most well-liked studies among the researchers correspond to their evolutionary field. Numerous threats addressed by data streams on clustering like limited time, memory and single scan data. In generic terms, data streams are defining an infinite sequence of the element data and evolve without prior knowledge number of the clusters. Many factors such as noise (outliers) have appeared periodically have a negative impact on data streams environment. The density-based technique has demonstrated to be an astonishing method in clustering data streams. It is computationally competent to generate arbitrary shape clusters and detect noise instantaneously. The number of the clusters are not required in advance to set as a parameter during the assessment begin. In contradictory, traditional density-based clustering is not relevant to conduct in data streams due to its own characteristics. Mostly all traditional density-based clustering can be extended to the updated version of algorithms to achieve the objective of data streams research. The idea is emphasizing on the density-based technique in the clustering process dominate the restrain from data streams nature. The objective of this paper intends a preliminary result on a density-based algorithm, named evoStream to explore outlier detection on medical data sets, heart failure clinical records and gene expression cancer RNA-seq. In due course, extensive evoStream later to develop to optimize the model to detect outlier in data streams.KeywordsData streamsOutlierDensity-basedevoStreamData setsEvaluation metrices