Análise de agrupamentos para dados espectrais

Abstract

Spectroscopy is the study that uses techniques to measure the spectrum of electromagnetic radiation, including visible light and radiofrequency, where we search for information such as chemical composition, temperature, density, mass, distance, luminosity, and relative motion using displacement measurements. In the case of Raman spectroscopy, we use a light diffusion process to obtain additional information about vibrations that increase the understanding of the fundamental molecular structure.

These methodologies provide a diversity of data, which will be modeled and analyzed using statistical and machine learning techniques. The spectroscopy data show high dimensionality and a strong presence of outliers that cause difficulties in clustering due to false positives in discovering new clusters.

For the study, a review of the literature on methods for grouping spectroscopy data that do not vary with time will be made. Thus, comparing models with the existing ones, and applying it to real data and simulated data.