1. Measurement of the Atmospheric Muon Neutrino Energy Spectrum with IceCube in the 79- and 86-String Configuration Tim Ruhe, Mathis Börner, Florian Scheriau, Martin Schmitz, TU Dortmund Flux Energy

2. 2 Outline  IceCube and atmospheric neutrinos  Event Selection  Spectrum Unfolding  Results  Summary and Outlook

3. 3 IceCube and atmospheric neutrinos Tim Ruhe | Statistische Methoden der Datenanalyse

4. 4 Data Preprocessing Cut on zenith angle > 86° Redcution of the data volume, BUT remaining background is significantly harder to reject Additional cuts:  Lepton velocity  Empty Hits  Truncated Energy (estimator)  Track length

5. 5 Feature Selection Stability Jaccard: Average over many sets of variables: Number of Variables Jaccard Index

6. 6 Training and Validation of a Random Forest  use an ensemble of simple decision trees  Obtain final classification as an average over all trees

7. 7 Random Forest Output  200 trees  5 Random Features per node  120,000 signal events  30,000 background events Forest Settings:

8. 8 Random Forest Output (IC79) Find a trade-off between background rejection and signal efficiency Place a cut at confidence >=0.92  212 neutrino candidates per day  66885 neutrino candidates in total  330±200 background muons

9. 9 Random Forest Output (IC86) 2-dimensional cut  289 neutrino candidates per day  92060 neutrino candidates in total  410±220 background muons

10. 10 Why unfold?  Muon production governed by stochastical processes  Energy losses on the way towards the detector  Finite resolution  Limited acceptance of the detector Inverse Problem, to be solved with TRUEE.

11. 11 Unfolding Input (1): Energy Correlation Input variables show good correlation with energy

12. 12 Unfolding Input (2): Background Distributions for IC79 Tim Ruhe | Statistische Methoden der Datenanalyse Background events are located at lower energies Therefore, they do not affect the highest energy bins

13. 13 Unfolding Input (3): Background Distributions for IC86 Background events are located at lower energies Therefore, they do not affect the highest energy bins

14. 14 Unfolding Input (3): Background Distributions for IC86

15. 15 Unfolding the spectrum TRUEE 3 energy dependent input variables TRUEE Clear deviation from an atmospheric only prediction

16. 16 More results Combining the spectra Unfolding different zenith bands

17. 17 Summary and Outlook Random Forest & MRMR TRUEE

18. 18 Backup Tim Ruhe | Statistische Methoden der Datenanalyse

19. 19 IC86 Precuts Tim Ruhe | Statistische Methoden der Datenanalyse

20. 20 IC79 Precuts Tim Ruhe | Statistische Methoden der Datenanalyse

21. 21 Comparison to other measurements Tim Ruhe | Statistische Methoden der Datenanalyse

22. 22 Relevance vs. Redundancy: MRMR (continuous case) Relevance: Redundancy: MRMR: or

23. 23 IC86 Confidence Distributions Tim Ruhe | Statistische Methoden der Datenanalyse

24. 24 Unfolding Input for IC86 Tim Ruhe | Statistische Methoden der Datenanalyse

25. 25 IC86 2D Cut Tim Ruhe | Statistische Methoden der Datenanalyse