Publication: Adquisición y análisis de información geolocalizable de una red social, apoyado en procesamiento paralelo
Loading...
Identifiers
Publication date
2011-06
Defense date
2011-06-16
Authors
Advisors
Tutors
Journal Title
Journal ISSN
Volume Title
Publisher
Impact
Export
Abstract
Este Proyecto Fin de Carrera estudia las diferentes áreas geográficas de incidencia de una persona, empresa, asociación, etcétera, a través de información susceptible de ser geolocalizada y que ha sido previamente obtenida de una red social. En concreto, la red social utilizada es Twitter y se emplean las relaciones de following (siguiendo) y follower (seguidor) para establecer los vínculos entre aquella persona (o más ampliamente, cuenta de Twitter que represente a cualquier entidad, sea cual sea esta) que será el origen del estudio, y aquellas otras sobre las que incide. Los datos sobre los que se ha orientado el estudio han sido tres: el campo ‘localización’ del perfil de usuario de Twitter, la etiqueta 'geo' que se encuentra adjunta a cualquier tweet (mensaje) que se publica y utilizar el texto de los tweets para realizar búsquedas sobre él. La cantidad de información que se deberá recoger de la red social es muy elevada, por lo que se propone para su tratamiento un enfoque que explote las bondades del procesamiento paralelo. Haciendo uso de una arquitectura paralela tipo clúster y utilizando un modelo de programación acorde a él, se estudiará el rendimiento de este método frente a una aproximación clásica no paralela. El procesamiento paralelo se ha desarrollado bajo un modelo maestro-esclavo, en el que el primero se encarga de crear una porción de trabajo que después asigna a un esclavo que se encuentre libre. El esclavo será el que procese estos datos y sea capaz de obtener resultados parciales relativos a esa asignación de trabajo. Finalmente, todos los esclavos devolverán los resultados parciales al maestro, quien calculará los resultados definitivos. __________________________________________________________________________________________________________________________
This Project aims to study the geographical areas in where a person, a company, an association, etc, has influence by using information capable of being localized and which has been previously gathered up from a social network. More specifically, the social network is Twitter, and its following and follower relationships are used to establish links between that person (or more generically, a Twitter account which represents to any entity, whichever it might be) who is the root of the study, and the other ones related to them. The study is focus on three pieces of data: the ‘location’ field of a Twitter user profile, the 'geo' tag attached to any tweet (message) published and by searching on the tweet text itself. The quantity of information that is required to be gathered up from the social network is very high, so for its processing it will be used a parallel computing approach. By using a parallel computing architecture such as a cluster and a parallel programming model, the performance of this method will be studied, and it will prove that it is better than a classic non-parallel approach. The parallel processing has been developed following a master-slave model: the first one is responsible for creating a piece of work which later will be assigned to an idle slave. The slave will process the piece of work and will get partial results related to this assignment. Finally, all the slaves will return their partial results them obtained to the master, who will calculate the definitive results.
This Project aims to study the geographical areas in where a person, a company, an association, etc, has influence by using information capable of being localized and which has been previously gathered up from a social network. More specifically, the social network is Twitter, and its following and follower relationships are used to establish links between that person (or more generically, a Twitter account which represents to any entity, whichever it might be) who is the root of the study, and the other ones related to them. The study is focus on three pieces of data: the ‘location’ field of a Twitter user profile, the 'geo' tag attached to any tweet (message) published and by searching on the tweet text itself. The quantity of information that is required to be gathered up from the social network is very high, so for its processing it will be used a parallel computing approach. By using a parallel computing architecture such as a cluster and a parallel programming model, the performance of this method will be studied, and it will prove that it is better than a classic non-parallel approach. The parallel processing has been developed following a master-slave model: the first one is responsible for creating a piece of work which later will be assigned to an idle slave. The slave will process the piece of work and will get partial results related to this assignment. Finally, all the slaves will return their partial results them obtained to the master, who will calculate the definitive results.
Description
Keywords
Diseño de software, Proceso en paralelo, Geolocalización, Redes sociales