| Name: | Description: | Size: | Format: | |
|---|---|---|---|---|
| 2.32 MB | Adobe PDF |
Authors
Advisor(s)
Abstract(s)
A extração de conhecimento de dados geolocalizados (geo-mining) tem contribuído
para a resolução de problemas existentes em diversos domínios tais como,
na saúde, meteorologia ou segurança, e.g., na identificação de locais de concentração
de crimes como meio de prevenção, e também no planeamento urbano ou
proteção civil, e.g., na elaboração de modelos de riscos de incêndio como suporte
ao planeamento das ações de busca e salvamento.
Neste trabalho desenvolve-se uma plataforma que a partir de dados geolocalizados
extrai conhecimento sob a forma de regras de associação espacial. As regras
extraídas são constituídas por relacionamentos espaciais, topológicos e de
distância. Os relacionamentos são obtidos entre entidades espaciais existentes na
superfície terrestre e elementos espaciais que caracterizam um fenómeno, e.g., um
foco de gripe aviária ou acidentes rodoviários, adicionados de um buffer espacial
de dimensão variável. A informação geoespacial utilizada foi obtida do projeto
colaborativo OpenStreetMap. Os dados OpenStreetMap são coletados a partir de
colaboração voluntária segundo o conceito Volunteered Geographic Information.
Um processo de extração de regras de associação produz, normalmente, um elevado
número de regras de associação o que dificulta a perceção de novo conhecimento.
Para abordar este problema é constituída uma geo-ontologia, com relacionamentos
espaciais detetados ‘a priori’, que originarão padrões conhecidos.
Estes relacionamentos são obtidos através da análise do dataset produzido. A
plataforma desenvolvida permite também integrar conhecimento do utilizador.
A metodologia utilizada para extração de regras de associação espacial permite
obter do espaço de pesquisa, um número de regras que não excede um valor
definido pelo utilizador.
Apresenta-se um estudo de caso sobre o problema da gripe aviária. As experiências
realizadas mostram resultados que seguem um padrão que foi confrontado,
com documentação especializada na área, que validam os resultados obtidos.
The extraction of geo-mining data has contributed towards the resolution of issues concerning a variety of domains, such as health, meteorology or security, for example, in identifying crime concentration sites, as a means of prevention and also in urban planning or civil protection, i.e., in preparing fire risk models as support towards the planning of search and rescue measures. This study focuses on the development of a platform that extracts knowledge in the form of rules of spatial association, taken from geo-mining data. The extracted rules consist of spatial, topological and distance relationships. The relationships are obtained from spatial entities that exist on the terrestrial surface and spatial elements that characterise a phenomenon, e.g., a spout of avian flu or road accidents, with the addition of a spatial buffer of variable size. The geospatial information used was obtained from a collaborative project - OpenStreetMap. The OpenStreetMap data is collected according to the concept of Volunteered Geographic Information, taken from voluntary cooperation. A given process of extracting rules of association, usually produces a high number of rules of association, which makes it difficult to perceive new knowledge. In order to address this issue, a geo-ontology is created, with spatial relationships detected beforehand, which will originate known patterns. These relationships are obtained through the analysis of the dataset produced. The developed platform also enables the integration of user knowledge. The methodology used for extracting rules of association enables one to obtain a series of rules, from the search area, that do not exceed a user-defined value. Shown here, is a case study on the problem of avian flu. The experiences conducted show results that follow a pattern, which were then compared against documentation from experts in this field, thus validating the results obtained.
The extraction of geo-mining data has contributed towards the resolution of issues concerning a variety of domains, such as health, meteorology or security, for example, in identifying crime concentration sites, as a means of prevention and also in urban planning or civil protection, i.e., in preparing fire risk models as support towards the planning of search and rescue measures. This study focuses on the development of a platform that extracts knowledge in the form of rules of spatial association, taken from geo-mining data. The extracted rules consist of spatial, topological and distance relationships. The relationships are obtained from spatial entities that exist on the terrestrial surface and spatial elements that characterise a phenomenon, e.g., a spout of avian flu or road accidents, with the addition of a spatial buffer of variable size. The geospatial information used was obtained from a collaborative project - OpenStreetMap. The OpenStreetMap data is collected according to the concept of Volunteered Geographic Information, taken from voluntary cooperation. A given process of extracting rules of association, usually produces a high number of rules of association, which makes it difficult to perceive new knowledge. In order to address this issue, a geo-ontology is created, with spatial relationships detected beforehand, which will originate known patterns. These relationships are obtained through the analysis of the dataset produced. The developed platform also enables the integration of user knowledge. The methodology used for extracting rules of association enables one to obtain a series of rules, from the search area, that do not exceed a user-defined value. Shown here, is a case study on the problem of avian flu. The experiences conducted show results that follow a pattern, which were then compared against documentation from experts in this field, thus validating the results obtained.
Description
Dissertação para obtenção do grau de Mestre em Engenharia Informática e de Computadores
Keywords
Bases de dados geográficas Geographic databases Geo-ontologias Geo-ontologies Data mining
Citation
PAREDES, António José da Silva - Extração de conhecimento em bases de dados geográficas com recurso a geo-ontologia. Lisboa: Instituto Superior de Engenharia de Lisboa, 2017. Dissertação de mestrado.
Publisher
Instituto Superior de Engenharia de Lisboa