Desenvolvimento de rotinas em Python para seleção de variáveis em dados espectroscópicos

Abstract

Bird sexing, crucial for poultry farming, domestic breeding, and species preservation, faces the challenge of identifying the sex in young birds, often without sexual dimorphism. Various techniques are used, each with its advantages and disadvantages. With this in mind, the present project addresses the gender identification in birds of the species bicudo, cockatiel, curió, and ring neck through the spectral analysis of infrared by Fourier transform – FTIR, and python language Random Forest algorithm. Integrating spectroscopy to detect subtle differences in feathers, the Random Forest is trained to distinguish between males and females. The success demonstrated in this approach, validated by external samples with 100% accuracy for the bicudos, highlights its potential for interdisciplinary applications and emphasizes the intersection between data analysis and applied physics.