Eye and skin color prediction for Brazilian population using single nucleotide polymorphisms ; Previsão da cor dos olhos e da pele para a população brasileiros utilizando polimorfismos de nucleotídeo único

When a crime is under investigation, especially when too many questions are unanswered, it is necessary to reduce the number of suspects to be able to solve the investigation. To reduce the number of suspects, any detail found at the crime scene is important, such as a strand of hair, DNA, or even a fingerprint. When the DNA found does not have the com plete information to be able to determine the identity of the suspect, some information can still be extracted from it, like the information of eye color or skin color. This work presents the application of Machine Learning algorithms, such as Random Forest, and Support Vector Machine to determine the pigmentation of the eye and skin using Single Nucleotide Polymorphisms (SNPs) from a DNA sample for forensics use. The follow ing chapters will present the necessary studies to investigate a solution for the proposed problem. Genetic and machine learning theoretical basis are presented, as well as related works, experiments, and results. Each dataset contains sixty-six SNPs and three classes: Blue, Intermediate, and Dark Brown are the classes related to eye color, and White, Inter mediate, and Brown are the classes related to skin color. 144 experiments were executed (72 for eye and 72 for skin classification), combining different approaches of feature se lection, class balanced, and classifiers to define the best solution. The data used for this study were collected from the Southern Brazilian population. The final results showed that 4 SNPs can be used to predict Blue and Dark Brown classes. For skin classification, 56 SNPs can be used when SMOTE is applied to balance the classes, but a further inves tigation is necessary to understand if the SMOTE is impacting the selection of the SNPs. Using 36 SNPs without class balance also achieved a close result. All the experiments had a bad performance for the Intermediate classes. For future work, a better investigation of intermediate colors is necessary. ; Quando um crime está sob investigação, especialmente quando ...