PREDICTING AND CLASSIFYING PACKET TRANSMISSION EFFICIENCY IN BIO-INSPIRED WIRELESS SENSOR NETWORKS
Ahmad Alkazzaz · VCU Scholars Compass (Virginia Commonwealth University) · 2014
Biological networks (specifically genetic regulatory networks) are known to be robust to various external perturbations. Bio-inspired wireless sensor networks (WSN) are known to be smart communication structures and have a have high packet transmission efficiency. In earlier work neural network models that correlate the average packet receival rates to the five topological features of the bio-inspired WSN were investigated. These features include the degree index, sink coverage, network density, hub node density, and motif index. In this thesis, an appropriate classification algorithm that works with these five features is investigated. The random forest algorithm is the best classification algorithm compared to other classification methods (APPENDIX B). In addition, a local weighted linear regression algorithm was created to predict the robustness of the network utilizing these five topological features.