Development of Wireless Sensor Network Congestion Detection Classifier Using Support Vector Machine

The interconnection of sensor nodes is deployed to monitor physical phenomenon known as wireless sensor network (WSN) . When an event occurs the congestion occurs in network near the nodes at sink. Support vector machine (SVM) is useful classification algorithm of machine learning .In this paper LibSVM(open source library for SVM) and SMO (sequential minimal optimization) algorithms are applied to build classifier for congestion detection into three levels low, medium and high using wireless sensor network traffic parameter data. The kernel function is used to map the non linearly separable data points into higher dimensions. The kernel used is radial basis function (RBF) as it has less computational complexity and it produces simple model. While choosing the model with RBF as kernel function, the parameters gamma (γ) and cost(c) plays important role. The aim of this paper is to find the good pair of (γ,c)values for sensor network data set. The grid search is performed with exponentially growing sequence of γ and c. The optimized model with gamma 23 and cost is 215 has been preferred for the sensor network traffic dataset based on percentage of accuracy and false positive rate.