Clústering Inteligente sobre Series Temporales
Loading...
Official URL
Full text at PDC
Publication date
2023
Authors
Advisors (or tutors)
Editors
Journal Title
Journal ISSN
Volume Title
Publisher
Citation
Abstract
Este trabajo de fin de máster está orientado al campo de la inteligencia artificial, concretamente a los algoritmos de aprendizaje automático. En particular, y dentro del aprendizaje automático no supervisado, se centra en el clustering de señales de series temporales, aplicación de especial interés en el contexto del Internet de las Cosas, donde una gran parte de los datos recolectados, especialmente aquellos provenientes de sensores, son de este tipo. En concreto, se presenta una versión del algoritmo TSK-means, un método propuesto en un artículo de investigación, cuya implementación no resulta trivial. Adicionalmente, se ha desarrollado una función generadora de conjuntos de datos personalizada, para disponer de conjuntos de entrenamiento con características concretas. Utilizando esta herramienta, realizamos un proceso de análisis y evaluación para identificar potenciales puntos de mejora. Fruto de esta evaluación, señalamos tres posibles problemas, para los que proponemos varias optimizaciones. Estos problemas están ligados al número de clústeres que el algoritmo busca obtener, y las optimizaciones desarrolladas aportan mayor flexibilidad al sistema, gracias a la posibilidad de variar este número durante la ejecución. En última instancia, el sistema evoluciona de un algoritmo independiente a una herramienta de exploración de datos.
This master thesis is oriented to the field of artificial intelligence, specifically to machine learning algorithms. In particular, and within unsupervised machine learning, it focuses on the clustering of time series signals. This is an application of special interest in the context of the Internet of Things, where a large part of the data collected, especially those coming from sensors, are of this type. Specifically, a version of the TSK-means algorithm, a method proposed in a research paper, whose implementation is not trivial, is presented. Additionally, a customized dataset generator function has been developed to provide training sets with specific characteristics. Using this tool, we performed an analysis and evaluation process to identify potential points of improvement. As a result of this evaluation, we point out three possible problems, for which we propose several optimizations. These problems are linked to the number of clusters that the algorithm seeks to obtain, and the optimizations developed provide greater flexibility to the system, thanks to the possibility of varying this number during execution. Ultimately, the system evolves from a stand-alone algorithm to a data exploration tool.
This master thesis is oriented to the field of artificial intelligence, specifically to machine learning algorithms. In particular, and within unsupervised machine learning, it focuses on the clustering of time series signals. This is an application of special interest in the context of the Internet of Things, where a large part of the data collected, especially those coming from sensors, are of this type. Specifically, a version of the TSK-means algorithm, a method proposed in a research paper, whose implementation is not trivial, is presented. Additionally, a customized dataset generator function has been developed to provide training sets with specific characteristics. Using this tool, we performed an analysis and evaluation process to identify potential points of improvement. As a result of this evaluation, we point out three possible problems, for which we propose several optimizations. These problems are linked to the number of clusters that the algorithm seeks to obtain, and the optimizations developed provide greater flexibility to the system, thanks to the possibility of varying this number during execution. Ultimately, the system evolves from a stand-alone algorithm to a data exploration tool.
Description
Trabajo de Fin de Máster en Internet de las Cosas, Facultad de Informática UCM, Departamento de Sistemas Informáticos y Computación, Curso 2022/2023