Metodología para la incorporación de un paquete de clasificación a la plataforma WEKA. Implementación utilizando un paquete para redes neuronales recurrentes
Cargando...
Fecha
Autores
Gallardo Segura, Alexy
Título de la revista
ISSN de la revista
Título del volumen
Editor
Universidad Central “Marta Abreu” de Las Villas
Resumen
Actualmente, las Redes Neuronales Artificiales constituyen un amplio campo de investigación. Estas redes se encuentran dentro de las técnicas conexionistas del campo de la Inteligencia Artificial y están orientadas a resolver problemas utilizando clasificación, regresión, reconocimiento de patrones, etc. El amplio uso de las Redes Neuronales Artificiales se refleja en el interés de las grandes compañías de software por incluirlas en sus principales paquetes matemáticos (por ejemplo MatLab, Wolfram Mathematica, IBM SPSS, etc.). Además algunas plataformas de aprendizaje automatizado, como la reconocida Weka, incluyen estos modelos en sus recursos. Particularmente las Redes Neuronales Recurrentes son altamente apropiadas para problemas de Bioinformática. En este trabajo se construye un paquete que contiene una implementación genérica de Redes Neuronales Recurrentes para la versión más reciente de la plataforma Weka. La principal contribución de este trabajo es el desarrollo de una metodología para la incorporación de nuevos paquetes a las versiones recientes de esta plataforma, explorando todas las vías posibles para ello. Finalmente, se evalúa la metodología e implementación utilizando un caso de estudio relacionado con un problema de Bioinformática, probando el funcionamiento correcto de la implementación y demostrando estadísticamente que la calidad del clasificador no es afectada durante la migración a la nueva versión de Weka.
Nowadays, Artificial Neural Networks are an open field for scientific research. These networks are within the connectionist techniques of the Artificial Intelligence field and are oriented to solve problems using classification, pattern recognition, regression, etc. The worldwide use of the Artificial Neural Networks is reflected on the interest of big software companies on include them within their main mathematical packages (such as MatLab, Wolfram Mathematica, IBM SPSS, etc.). Also some other platforms for machine learning, such as the well-known Weka, include these models in their resources. Particularly, Recurrent Neural Networks are highly suitable for Bioinformatics problems. In this work we build a package containing an implementation of generic Recurrent Neural Networks for the recent version of Weka platform. As the main contribution we elaborate a methodology for the incorporation of new packages to the recent version of this platform, exploring all possible ways. Finally, we validate the implementation using a case of study related to a Bioinformatics problem, showing the correct functioning of the implementation and statistically demonstrating that the classifier performance has not been affected during the migration to the new version of Weka.
Nowadays, Artificial Neural Networks are an open field for scientific research. These networks are within the connectionist techniques of the Artificial Intelligence field and are oriented to solve problems using classification, pattern recognition, regression, etc. The worldwide use of the Artificial Neural Networks is reflected on the interest of big software companies on include them within their main mathematical packages (such as MatLab, Wolfram Mathematica, IBM SPSS, etc.). Also some other platforms for machine learning, such as the well-known Weka, include these models in their resources. Particularly, Recurrent Neural Networks are highly suitable for Bioinformatics problems. In this work we build a package containing an implementation of generic Recurrent Neural Networks for the recent version of Weka platform. As the main contribution we elaborate a methodology for the incorporation of new packages to the recent version of this platform, exploring all possible ways. Finally, we validate the implementation using a case of study related to a Bioinformatics problem, showing the correct functioning of the implementation and statistically demonstrating that the classifier performance has not been affected during the migration to the new version of Weka.