Hybrid lion and exponential PSO-based metaheuristic clustering approach for efficient dynamic data stream management

In dynamic data stream environment, the problem related to the exploration of big data within the real time scenario cannot be addressed through the tracking of each individual historic data even though it is highly memory expensive. Thus, a data stream clustering method is essential for exploring and storing the potential amount of information from the historical data determined in a single pass. The dynamic algorithms developed for clustering need to satisfy the two requirements of concept drift and concept evolution. These dynamic algorithms need to handle the change in the association between the object attributes that are existing within each individual clusters. In this paper, A Hybrid Lion and Exponential PSO-based Metaheuristic Clustering Approach (HLEPSOMCA) is proposed for satisfying the requirements of concept drift and concept evolution during efficient dynamic data stream management. This Metaheuristic Clustering Approach is proposed with the properties of good scalability and minimized number of parameters with respect to the number of clusters and high dimensional data determined from the dataset. It adopted different methods of stochastic optimization and deterministic clustering techniques for centring the clusters in an optimal manner. It further adopted density clustering strategies for determining micro clusters, such that Lion and Exponential PSO can be adopted in the initialization phase for maximizing the performance. The experimental results of this HLEPSOMCA approach with respect to KDD-99 dataset confirmed that the purity achieved by the proposed HLEPSOMCA scheme is improved on an average by 13.24%, better than the bassline approaches used for comparison.