Identification of cosmic ray electrons and positrons by neural networks

Abstract A data analysis based on artificial neural network classifiers has been done to identify cosmic ray electrons and positrons detected with the balloon-borne NMSU/Wizard-TS93 experiment. The information is provided by two ancillary and independent particle detectors: a transition radiation detector and a silicon-tungsten imaging calorimeter. Electrons and positrons measured during the flight have been identified with background rejection factors of 80 ± 3 and 500 ± 37 at signal efficiencies of 72 ± 3% and 86 ± 2% for the transition radiation detector and silicon-tungsten imaging calorimeter, respectively. The ability of the artificial neural network classifiers to perform a careful multidimensional analysis surpasses the results achieved by conventional methods.