A Comparative Study of BDT and DNN Algorithms for Higgs Boson Prediction

Abstract

The Higgs boson, also known as the "God Particle," is responsible for giving mass to elementary particles. Detecting and studying its production remains a major challenge in particle physics. In this study, we use deep neural networks and decision-boosted trees to identify the decay of the Higgs boson into four leptons. Our dataset includes millions of simulated collision events from the Large Hadron Collider. Results show that decision-boosted tree models are highly effective in recognizing complex patterns, improving the accuracy of Higgs boson detection. We evaluate model performance using key metrics such as ROC-AUC curves, Cross-Validation, background-to-noise ratio, and score distribution analysis. Our findings offer a strong framework for advancing Higgs boson research in high-energy physics.