Técnicas de Machine Learning y desarrollo de modelos predictivos aplicados a la antropología forense

Abstract

En este trabajo se estudia la aplicación de las técnicas de Machine Learning en el desarrollo de modelos de clasificación para la estimación del sexo de restos óseos humanos, dentro de un contexto forense y bioarqueológico. Estas técnicas permiten desarrollar algoritmos que puede aprender y esquematizar propiedades y patrones estructurales subyacentes de los datos, pudiendo utilizarse esta información para entender y predecir fenómenos específicos. Por tanto, dada la importancia que tiene una estimación fiable del sexo y la utilidad de las técnicas de Machine Learning en el desarrollo de modelos de clasificación para la predicción de variables categóricas. Lo que se propone en este trabajo es la aplicación de algoritmos estadísticos clásicos y de Machine Learning para la construcción de modelos predictivos para la estimación del sexo a partir de estudios métricos de estos restos. Para ello, en primer lugar se realizó una revisión bibliográfica de las técnicas de Machine Learning y de estadística clásica utilizadas en el área de la antropología forense. Obteniéndose un texto donde se definen y comparan estos métodos. Una vez hecho hecho se evaluaron y se escogieron los que podían adaptarse mejor a nuestro problema. Por último, se obtuvieron tres modelos de clasificación con los métodos de regresión logística, redes neuronales artificiales y Random Forest, utilizando el software R. Los resultados obtenidos fueron una serie de potentes modelos predictivos. Por lo que se puede decir que las técnicas de Machine Learning son una prometedora alternativa a los métodos clásicos de clasificación.

This paper studies the application of Machine Learning techniques in the development of models for estimating the sex of human beings, within a forensic and bioarchaeological context. These techniques allow the development of algorithms that can learn and schematize properties and underlying structural patterns of the data, and can be used for specific information and predict specific phenomena. Therefore, given the importance of a reliable estimation of sex and the usefulness of Machine Learning techniques in the development of classification models for the prediction of categorical variables. What is offered in this work is the application of classical algorithms and Machine Learning for the construction of predictive models for sex estimation from metric studies of these remains. For this, in the first place, a bibliographic review of the Machine Learning and Statistics Classics techniques in the area of forensic anthropology was carried out. Obtaining a text where these methods are defined and compared. Once done, we evaluated and chose those who stayed better to our problem. Finally, three classification models were obtained with the methods of logistic regression, artificial neural networks and Random Forest, using the R software. The results were a series of powerful predictive models. Therefore, it can be said that Machine Learning techniques are a primising alternative to classical classification methods.

This paper studies the application of Machine Learning techniques in the development of models for estimating the sex of human beings, within a forensic and bioarchaeological context. These techniques allow the development of algorithms that can learn and schematize properties and underlying structural patterns of the data, and can be used for specific information and predict specific phenomena. Therefore, given the importance of a reliable estimation of sex and the usefulness of Machine Learning techniques in the development of classification models for the prediction of categorical variables. What is offered in this work is the application of classical algorithms and Machine Learning for the construction of predictive models for sex estimation from metric studies of these remains. For this, in the first place, a bibliographic review of the Machine Learning and Statistics Classics techniques in the area of forensic anthropology was carried out. Obtaining a text where these methods are defined and compared. Once done, we evaluated and chose those who stayed better to our problem. Finally, three classification models were obtained with the methods of logistic regression, artificial neural networks and Random Forest, using the R software. The results were a series of powerful predictive models. Therefore, it can be said that Machine Learning techniques are a primising alternative to classical classification methods.