Detecting Neutral Nodes in a Network of Heterogeneous Agent Based System

Abstract

Software agents with autonomous interaction, negotiation and learning capabilities can be considered as a social network. Central problems in these social networks are: (1) investigating the network to find hot spots (nodes that actively participate in the expansion of the net both physically and functionally); (2) detecting violating nodes, i.e. nodes that have violated certain network policies. Key difference between human and agent based social nets are heterogeneity of agents and having various instances for an agent type. In this paper we propose a technique that can identify neutral nodes, i.e. nodes that can never be considered as a violating or interesting node because it shows similar behavior in all of its interactions. Given the communication protocol and by extracting communication rules, the neutrality for the agent is found in two levels, namely, inter-type and intra-type interactions. The neutrality is found in two steps by giving the network topology and rules of negotiation among agents, and the communication protocol and the extracted communication rules. These two steps guarantee the similarity in frequencies and behaviors of the agents. A direct advantage of this method is reducing the computational complexity for applying analysis techniques and examining all agents to detect a certain property. The techniques are verified through a case study in online auctions. Two additional application areas which widely use agent-based systems are discussed for utilization of the proposed method: e-commerce and e-learning.