Cluster Scanning: a novel approach to resonance searches

• DOI: 10.48550/arxiv.2402.17714
• Publication Date: 2024-02-27
• Authors: Ivan Oleksiyuk; J. A. Raine; M. Krämer; Slava Voloshynovskiy; T. Golling
• Keywords: FOS: Physical sciences; QC770-798; 530; info:eu-repo/classification/ddc/500.2; High Energy Physics - Experiment; High Energy Physics - Phenomenology; High Energy Physics - Experiment (hep-ex); High Energy Physics - Phenomenology (hep-ph); Jets and Jet Substructure; Nuclear and particle physics. Atomic energy. Radioactivity; Physics - Data Analysis, Statistics and Probability; info:eu-repo/classification/ddc/025.063; info:eu-repo/classification/ddc/530; Specific BSM Phenomenology; Data Analysis, Statistics and Probability (physics.data-an)

We propose a new model-independent method for physics searches called Cluster Scanning. It uses the k-means algorithm to perform clustering in space of low-level event or jet observables, and separates potentially anomalous clusters construct signal-enriched region. The invariant mass spectra these two regions are then used determine whether resonant signal is present. A pseudo-analysis on LHC Olympics dataset with $Z'$ resonance shows that Scanning outperforms widely 4-parameter functional background fitting procedures, reducing number events needed reach $3\sigma$ significant access by factor 0.61. Emphasis placed speed method, which allows test statistic be calibrated synthetic data.