Management Of Sensor Network Using Dynamic Subgraph Mining

Sensor Networks are composed of low-power distributed devices called sensors which are capable of performing a set of activities such as sensing data, processing and communication. Although individual sensor's processing power is limited, a network of a set of sensors is capable of completing a task - big or small - quite efficiently. However, failure of sensor networks results in the need for managing these networks efficiently so that whole system works properly. One of the requirements for efficient management is to identify the relevant information of the desired set of sensors quickly. This is the topic of this thesis. We use a frequent dynamic subgraph mining algorithm to identify necessary communication patterns created by these logically related sensors. The entire process is known as Sensor mining. The sensor miner was successfully implemented and tested against different sensor network graphs, resulting in the efficient identification of desired set of sensors.