Abstract

Abstract. The development of new reliable data processing and mining methods based on the synergy between cloud computing and the multi-agent paradigm is of great importance for contemporary and future software systems. Cloud computing provides huge volumes of data and computational resources, whereas the agents make the system components more autonomous, cooperative, and intelligent. This creates the need and gives a very good basis for the development of data analysis, processing, and mining methods to enhance the new agent-based cloud computing (ABCC) architecture. Ad-hoc networks of virtual agents are created in the ABCC architecture to support the dynamic functionality of provided services, and data processing methods are very important at the input data processing and network parameter estimation stage. In this study, we present a decentralized kernel-density-based clustering algorithm that fits with the general architecture of ABCC systems. We conduct several experiments to demonstrate the capabilities of the new approach and analyse its efficiency.