trabajo final de especialización.page.titleprefix Detección automática de anomalías en logs: una revisión visual del estado del arte
Loading...
Date
2022-10-05
Authors
Journal Title
Journal ISSN
Volume Title
Publisher
Abstract
"Con el objetivo de elaborar una revisión visual del estado del arte de la detección automática de anomalías en logs, se recopilaron 20 trabajos (publicados entre 2009 y 2021) centrados en dicha línea de investigación. 17 de ellos son trabajos experimentales y 3 son revisiones. Los trabajos fueron procesados para extraer información descriptiva asociada a la publicación y el contenido. Se prestó particular atención a las categorías de aprendizaje y los modelos de aprendizaje automático entrenados en los trabajos para detectar anomalías. También se prestó atención a la interacción entre los trabajos, a través de las citas. Con la información extraída, se construyó una base de datos de 3 tablas describiendo los
autores, los trabajos y sus interacciones.
Con los datos, se construyó un notebook de visualizaciones en ObservableHQ. Estas son útiles para obtener una idea inicial de los países del mundo más involucrados en esta investigación, los trabajos más influyentes y los modelos de aprendizaje más utilizados. Dado que el tamaño muestral de trabajos utilizados aquí es pequeño, no se plantea que las tendencias observadas por las visualizaciones sean representativas de las tendencias reales. Sin embargo, con un mayor tamaño muestral, esta revisión visual podría ser útil para resumir información importante sobre el estado del arte de este tema, de
manera que un lector no especializado pueda identificar rápidamente trabajos de interés para sus requerimientos específicos."
"With the objective of developing a visual survey of the state of the art of automatic anomaly detection in logs, 20 papers (published between 2009 and 2021) centered on this topic were gathered. 17 of them are experimental projects, and 3 are surveys. The papers were processed in order to extract descriptive information associated with the publication and the content. Particular attention was paid to the learning categories and the machine learning models trained to detect anomalies. The interactions between the papers was also noted, by way of their citations. With the information gathered, a database with 3 tables was constructed, describing the authors, the papers and their interactions. With this data, a notebook of visualizations was built in ObservableHQ. These visualizations are useful to obtain an initial understanding of the countries most involved in this research, the most influential papers, and the most commonly used learning models. Given that the sample size of papers taken into account here is small, the tendencies observed by these visualizations are not considered to be representative of reality. However, with a larger sample size, this visual survey could be useful to summarize important information about the state of the art of this topic, so that a non-specialized reader can quickly identify papers of interest for their specific requirements."
"With the objective of developing a visual survey of the state of the art of automatic anomaly detection in logs, 20 papers (published between 2009 and 2021) centered on this topic were gathered. 17 of them are experimental projects, and 3 are surveys. The papers were processed in order to extract descriptive information associated with the publication and the content. Particular attention was paid to the learning categories and the machine learning models trained to detect anomalies. The interactions between the papers was also noted, by way of their citations. With the information gathered, a database with 3 tables was constructed, describing the authors, the papers and their interactions. With this data, a notebook of visualizations was built in ObservableHQ. These visualizations are useful to obtain an initial understanding of the countries most involved in this research, the most influential papers, and the most commonly used learning models. Given that the sample size of papers taken into account here is small, the tendencies observed by these visualizations are not considered to be representative of reality. However, with a larger sample size, this visual survey could be useful to summarize important information about the state of the art of this topic, so that a non-specialized reader can quickly identify papers of interest for their specific requirements."
Description
Keywords
APRENDIZAJE AUTOMATICO, VISUALIZACION