Identificación De Patrones De Trayectorias Vehiculares Usando El Algoritmo K-Means
No hay miniatura disponible
Fecha
2017-07-20
Autores
Título de la revista
ISSN de la revista
Título del volumen
Editor
Universidad de Guayaquil.Facultad de Ciencias Matematicas y Fisicas.Carrera en Ingenieria en Sistemas Computacionales
Resumen
Esta investigación se centra en el estudio del algoritmo k-means como medio de agrupación de información con el objetivo de poder conocer su funcionamiento y procesamiento de información. De la información obtenida se diseñó los experimentos que a su vez finalizó con una comparación de los procesos de tres variantes del algoritmo. Estas variantes son: k-means++, k-means + elbow y k-means + pso, que han ido apareciendo con el tiempo según las necesidades de superar al original. Se usó la información de tres bases de datos la cuales son: la base de california, plt y t-drive, las mismas que contienen datos de trayectorias vehiculares. Como resultado se evidenció las ventajas de uso de cada algoritmo utilizando como métricas de comparación: margen de error, distancia intra-clúster, distancia inter-clúster, tiempo de ejecución y silhouette. Se utilizó la observacion como medio de obtención del conocimientos así como la revisión documental. A su vez en base a estas métricas se obtuvieron resultados que permitió la identificación de patrones para las trayectorias vehiculares que se estudian como es el tráfico vehicular, falta de señalización vial, entre otras.
This research focuses on the study of the k-means algorithm as a means of grouping information with the objective of knowing its operation and information processing. From the information obtained we designed the experiments which ended with a comparison of the processes of three variants of the algorithm. These variants are: k-means ++, k-means + elbow and k-means + pso, which have been emerging over the time according to the needs of overcoming the original one. Information from three databases was used: the base of California, plt and t-drive, which are the same ones that contain data of vehicular trajectories. As a result, there were found advantages of using each algorithm obtained by using comparison metrics: margin of error, intra-cluster distance, inter-cluster distance, time and silhouette. Also, observation was used as a means of obtaining knowledge as well as documentary review. In turn, based on these metrics, results were obtained for allowing the identification of patterns for vehicular trajectories studied such as vehicular traffic, lack of road signs, among others
This research focuses on the study of the k-means algorithm as a means of grouping information with the objective of knowing its operation and information processing. From the information obtained we designed the experiments which ended with a comparison of the processes of three variants of the algorithm. These variants are: k-means ++, k-means + elbow and k-means + pso, which have been emerging over the time according to the needs of overcoming the original one. Information from three databases was used: the base of California, plt and t-drive, which are the same ones that contain data of vehicular trajectories. As a result, there were found advantages of using each algorithm obtained by using comparison metrics: margin of error, intra-cluster distance, inter-cluster distance, time and silhouette. Also, observation was used as a means of obtaining knowledge as well as documentary review. In turn, based on these metrics, results were obtained for allowing the identification of patterns for vehicular trajectories studied such as vehicular traffic, lack of road signs, among others
Descripción
ADOBE
Palabras clave
ALGORITMO, K-MEANS