Data mining on ice

Tim Ruhe (TU Dortmund),
Prof. Dr. Katharina Morik (TU Dortmund)

Abstract


IceCube is a cubic kilometer neutrino telescope located at the geographic South

Pole. The detector consists of 5160 Digital Optical Modules (DOMs) installed on

86 vertical cables (strings). The main scientific goal of IceCube is the detection of

neutrinos from extragalactic sources like Active Galactic Nuclei and Gamma Ray

Bursts. For the detection of such sources however, a detailed understanding of the

atmospheric neutrino spectrum is crucial. These atmospheric neutrinos are produced

in cosmic ray air showers. The detection of atmospheric neutrinos is hindered by

atmospheric muons entering the detector from above. A first rejection of background

can be achieved by using the earth as a muon shield and selecting upward going

tracks only. A small fraction of atmospheric muons however, is still misreconstructed

as upgoing. Due to the small interaction cross section of neutrinos the number of

misreconstructed muons exceeds the number of actual neutrino events by roughly

three orders of magnitude at the analysis level. The low signal to background ratio makes this task well suited for a

study within the scope of machine learning. Detailed studies have been carried out in

order to achieve the best possible separation of signal and background. Our studies

extend from detailed parameter selection and parameter construction to detailed

investigations of diverse classifiers. Monte Carlo simulations as well as real data

were used.

Slides in PDF format

Paper ID: O24