A carregar...
Publicação
Modelação dos dados abertos de bike-sharing de Bruxelas (Villo) através de modelos de regressão espacial
| Nome: | Descrição: | Tamanho: | Formato: | |
|---|---|---|---|---|
| 11.47 MB | Adobe PDF |
Autores
Orientador(es)
Resumo(s)
O ciclismo utilitário tem despertado renovado interesse nos últimos anos devido ao reconhecido potencial
deste modo de transporte na redução ao da pegada energética e da poluição, ao nas cidades. A etiqueta de
meio de transporte sustentável é justificada, para além dos fatores citados, pelo reduzido espaço que a
bicicleta ocupa em comparação¸ ao com o automóvel, assim como pela pegada ecológica de toda a cadeia
de produção. Neste contexto, várias cidades a nível mundial têm levado a cabo iniciativas para incitar a
utilização quotidiana da bicicleta por parte dos habitantes. De entre estas iniciativas, os sistemas públicos
de bicicletas partilhadas (bike-sharing) são talvez o melhor exemplo dada a sua universalidade. Os sistemas
de bicicletas partilhadas além de possibilitarem a utilização da bicicleta sem necessidade de aquisição, da
mesma, são também uma fonte de dados valiosa, que pode ser usada para melhorar os sistemas existentes
ou para levar a cabo estudos de investigação sobre diferentes temas ligados à mobilidade e à forma como
os cidadãos se deslocam e interagem com o meio envolvente.
Neste trabalho, exploram-se os dados abertos do sistema público de partilha de bicicletas de Bruxelas
(Villo), com vista a elaborar um modelo de regressão espacial que permita estimar o número de viagens
em bicicleta. Procura-se entender os fatores que influenciam a mobilidade ao longo do dia e são analisados
os limites e potencialidades do modelo, nomeadamente no que diz respeito à aplicação a outros contextos
e à generalização para todas as viagens em bicicleta.
Neste estudo começa-se por realizar um modelo de regressão linear exploratório, que permite identificar
um conjunto de variáveis potencialmente explicativas do número de viagens em bicicleta nas estações.
No final parametriza-se um modelo de regressão geograficamente ponderado com distribuição de Poisson,
com as variáveis identificadas previamente.
Os resultados do modelo final sugerem que as relações entre as variáveis analisadas apresentam uma
grande variação espacial e temporal. O modelo permite além disso a revelação de relações locais escondidas,
que se contextualizadas com outros estudos, permitem um melhor conhecimento dos fatores que
afetam as viagens em bicicleta.
In recent years there has been a renewed interest in utilitarian cycling due to its recognized potential in the reduction of energy consumption and pollution in the cities. In addition to the above-mentioned factors, the attached label of sustainable means of transport is justified by the little space occupied by bicycles compared to cars, as well as by the smaller ecological impact of the entire production chain. Within this context, several cities around the world have been launching different initiatives of bike promotion. Among these initiatives, bike-sharing systems are perhaps the best example given their universality. Bikesharing systems allow users to ride a bicycle between two places (not necessarily different places) without the need of having a bike of one’s own. These systems generate a big amount of data which can then be used to improve the systems themselves or to carry out research studies on different subjects related to mobility and the way how citizens move around and interact with the surrounding environment. In this paper, the open data automatically generated by the Brussel’s bike-sharing system (Villo) is explored, aiming to elaborate a spatial regression model to estimate the number of bicycle trips at stations. The main goal of the modelling process is to understand the underlying factors influencing mobility patterns throughout the day. The weaknesses and strengths of the model approach are analyzed, especially in what regards its implementation in other geographic contexts and the possibilities of generalization for all bicycle trips. The first steps of the modelling process consist in setting up an exploratory Ordinary Least Squares model (OLS) in order to identify the potential explanatory variables among those available in the database. In the end, a GeographicallyWeighted Poisson Regression model (GWRP) is parametrized using the previous identified variables. The results suggest the relationships between the dependent and independent variables are complex and spatially varying. Furthermore, the results show hidden patterns that enable further local investigation on these relationships.
In recent years there has been a renewed interest in utilitarian cycling due to its recognized potential in the reduction of energy consumption and pollution in the cities. In addition to the above-mentioned factors, the attached label of sustainable means of transport is justified by the little space occupied by bicycles compared to cars, as well as by the smaller ecological impact of the entire production chain. Within this context, several cities around the world have been launching different initiatives of bike promotion. Among these initiatives, bike-sharing systems are perhaps the best example given their universality. Bikesharing systems allow users to ride a bicycle between two places (not necessarily different places) without the need of having a bike of one’s own. These systems generate a big amount of data which can then be used to improve the systems themselves or to carry out research studies on different subjects related to mobility and the way how citizens move around and interact with the surrounding environment. In this paper, the open data automatically generated by the Brussel’s bike-sharing system (Villo) is explored, aiming to elaborate a spatial regression model to estimate the number of bicycle trips at stations. The main goal of the modelling process is to understand the underlying factors influencing mobility patterns throughout the day. The weaknesses and strengths of the model approach are analyzed, especially in what regards its implementation in other geographic contexts and the possibilities of generalization for all bicycle trips. The first steps of the modelling process consist in setting up an exploratory Ordinary Least Squares model (OLS) in order to identify the potential explanatory variables among those available in the database. In the end, a GeographicallyWeighted Poisson Regression model (GWRP) is parametrized using the previous identified variables. The results suggest the relationships between the dependent and independent variables are complex and spatially varying. Furthermore, the results show hidden patterns that enable further local investigation on these relationships.
Descrição
Dissertation presented as the partial requirement for obtaining a Master's degree in Geographic Information Systems and Science
Palavras-chave
Bike-sharing Ciclismo utilitário Mobilidade Estatística espacial Modelação Regressão espacial Utility cycling Mobility Spatial statistics Modelling Spatial regression