Data mining classification of cars based on GPS shadows in Forward Scatter Radar systems

The goal of this paper is to introduce a new concept - using Data Mining approach for radar classification of vehicles, based on their GPS shadow signals detected in a GPS L1-based Forward Scatter Radar (FSR) system. Real data is used for the current experiments, recording the GPS radio shadows of several moving vehicles with commercial non-professional GPS equipment. The records are further processed in MATLAB computational environment in order to obtain the estimated parameters of the GPS shadows. Data Mining approach is then implemented for classifying the cars, comparing classifiers generated with different classification algorithms - a decision tree, a neural network, and a Bayesian classifier, for two different variants of the class variable, taking three or two possible values. The best results are achieved for the two values of the class variable and the best performing classifiers are the neural networks.