Node Localization in Wireless Sensor Networks Using the M5P Tree and SMOreg Algorithms

Abstract

In Wireless Sensor Networks (WSN), Node Localization is of great importance for location aware services. In this paper we propose the use of Time of Arrival (TOA) information with two popular machine learning algorithms M5 tree Model (M5P) and Sequential Minimal Optimization for Regression (SMOreg) for more accurate node localization in WSN. In this paper we also applied the same node localization problem to two previously used artificial neural network models- Multi Layer Perceptron (MLP) and Radial Basis Function (RBF) Network. After that a comparative analysis between all selected algorithms has been made. Simulation results show the superiority of M5P and SMOreg over MLP and RBFN in high noise conditions in terms of root mean square error. At last a comparative analysis between the two new proposed algorithms was made by changing the number of training nodes. Results show that initially the performance of SMOreg is better but there is no improvement in its performance with increasing training samples. On the other hand M5P's performance can be made better by train it with more number of samples.