An Efficient Simulated Annealing based Constrained Optimization Approach for Outlier Detection Mechanism in RFID-Sensor Integrated MANET

Abstract

Designing an outlier detection process for unknown resources is a challenging task. It may contain resourceful or resource-constrained devices. In this work, a multi-regional and multi-layered outlier detection process is proposed. Proposed approach implements MAC, routing and application layer outlier detection processes in three different regions. These regions are designed with priority of resources and importance of stakeholder taken into considerations. Similar outlier processed with different datasets is used for outlier detection in this multi-region invigilator architecture. Proposed architecture is verified through internal, external and performance based indices. Simulation results shows the cluster stability in process of data formalization and outlier detection. Internal and external indices shows that the maximum stability is possible with 5 to 500 nodes and 26 clusters for small scale network, 500 to 3000 nodes with 41 clusters for medium scale network and 3000 to 6000 nodes network with 54 clusters for large scale network.