Publication: Aplicación de sistemas de consenso sobre modelos de Machine Learning
Loading...
Official URL
Full text at PDC
Publication Date
2019
Advisors (or tutors)
Editors
Journal Title
Journal ISSN
Volume Title
Publisher
Citations
Exportar
Abstract
En este trabajo se investigan y comparan distintos algoritmos de consenso aplicados entre modelos de Machine Learning. Más concretamente, se aplican algoritmos de consenso basados en la metodología Delphi sobre cuatro datasets con diferentes características. Primero se seleccionan los datasets y se aplica una limpieza sobre los mismos, y a continuación se generan diversos modelos de clasificación mediante algoritmos de Machine Learning (K-Nearest Neighbors, Máquinas de Soporte Vectorial, Bosques Aleatorios, Árboles de decisión, Redes Neuronales y Regresión Logística). Junto a estos modelos se aplican los algoritmos de consenso ideados en base al método Delphi, analizando los resultados obtenidos con los mismos.
Además, se incorpora una interfaz gráfica para observar los resultados de forma mas sencilla y poder aplicar distintos ejemplos individuales de forma manual sobre los algoritmos de consenso y los modelos de clasificación deseados.
In this project various consensus algorithms are investigated, compared and applied to real world problems. The consensus algorithms are based on the Delphi method and applied with four different datasets. Firstly, the datasets are selected and cleaned, and then Machine Learning algorithms are used to create different classification models (KNearest Neighbor, Support Vector Machines, Random Forest, Decision Trees, Neural Networks and Logistic Regression).Along these models, the mentioned consensus algorithms created with our own approach are applied. Finally, the results are analyzed, reaching a conclusion about the use of meta-classifiers instead of the Machine Learning models alone. Also and user interface is implemented so the results and the process can be followed by the users, making it easier for anyone to apply the developed methods. The accuracy of the application of those algorithms can be compared to the users’ models, so the user can decide wheter to use or not the algorithms with their particular problems.
In this project various consensus algorithms are investigated, compared and applied to real world problems. The consensus algorithms are based on the Delphi method and applied with four different datasets. Firstly, the datasets are selected and cleaned, and then Machine Learning algorithms are used to create different classification models (KNearest Neighbor, Support Vector Machines, Random Forest, Decision Trees, Neural Networks and Logistic Regression).Along these models, the mentioned consensus algorithms created with our own approach are applied. Finally, the results are analyzed, reaching a conclusion about the use of meta-classifiers instead of the Machine Learning models alone. Also and user interface is implemented so the results and the process can be followed by the users, making it easier for anyone to apply the developed methods. The accuracy of the application of those algorithms can be compared to the users’ models, so the user can decide wheter to use or not the algorithms with their particular problems.
Description
Trabajo de Fin de Grado en Ingeniería de Computadores, Facultad de Informática UCM, Departamento de Ingeniería del Software e Inteligencia Artificial, Curso 2018/2019.