Development of an Event Detection System Based on a Wireless Sensor Network
Research Paper.doc (7.147Mb)
MetadataShow full item record
The focus of this research is to develop an event detection system for groundwater flooding. The groundwater monitoring infrastructure is based on a Wireless Sensor Network, which can often lead to corrupted data values. The event detection system seeks to develop an algorithm to identify these corrupted data values and flag them as a false event to the user. It aims to maintain a high accuracy of true events, while keeping the occurrence of a false event to a minimum. The algorithm is based on threshold classification technique to identify the event and a statistical modified Z Score calculation to determine if the event is true or false. I ran the algorithm with an historic flood event and measured the accuracy of the results with a ROC curve. Results showed a 95% accuracy level but it also identified 8 out of 160 true events as false, which leads me to recommend a further analysis into the threshold that defined the algorithm. The technique would also need to be tested on larger dataset to validate this accuracy.