Identifying Patterns for Urban Planning through Neural Networks

F.J. Abarca-Álvarez (2011). Identification of Patterns for Urban Planning using Neural Networks. Towards the Ordinance-Network.

Serie Geográfica

F.J. Abarca-Álvarez (2011). Identification of Patterns for Urban Planning using Neural Networks. Towards the Ordinance-Network.
Serie Geográfica

The Ordinance, as a link between the urban project and architecture, must adopt commitments with both purely urban concepts and more architectural ones, establishing patterns of relationship or network between them. As opposed to the stationary nature of conventional urban ordinances, which are usually out of date at the very moment of their formulation, we explore a way of working based on the feedback of the normative body through the dynamic introduction of both pre-existences and the transformations generated within the framework of the ordinance itself. We will call these network or relational patterns Ordinance-network. The adaptation of each proposal to the Ordinance-network will develop a normative framework capable of evolving that will be generated from the global as opposed to the particular and accidental. It will therefore be necessary to search for coherences that are formalised in patterns that group together and show the most representative variables of the selected area.

By means of the objective numerical representation of both the variables that a priori are considered valuable or representative and those whose interest is unknown, and their application by means of an artificial neural network with unsupervised and competitive learning of the self-organising map or SOM (Self-Organising Map) type, and more specifically the Kohonen network, the underlying structure of these variables can be discovered and represented in a comprehensible way.

The obtained results are easily interpretable, allowing to recognize the grouping in the shape of patterns of architectural or urban objects represented, being allowed then a simple verification of the integration of a new building or object to the set of patterns that compose the network ordinance. These pattterns are called Ecotypes. The new objects accepted by this standard will form part of this standard, being integrated in the body of a new neural network, achieving this way a continued feedback of the ordinance. The proposed method is conformed as a decision-making aid method and as a assissance tool to the project of an ordinance. As a verification of the method, an implementation of an network ordinance is proposed in the historical center of Santa Fe in the province of Granada, comparing the obtained results with the in effecto planning represents.

The Ordinance, as a link between the urban project and architecture, must adopt commitments with both purely urban concepts and more architectural ones, establishing patterns of relationship or network between them. As opposed to the stationary nature of conventional urban ordinances, which are usually out of date at the very moment of their formulation, we explore a way of working based on the feedback of the normative body through the dynamic introduction of both pre-existences and the transformations generated within the framework of the ordinance itself. We will call these network or relational patterns Ordinance-network. The adaptation of each proposal to the Ordinance-network will develop a normative framework capable of evolving that will be generated from the global as opposed to the particular and accidental. It will therefore be necessary to search for coherences that are formalised in patterns that group together and show the most representative variables of the selected area.

By means of the objective numerical representation of both the variables that a priori are considered valuable or representative and those whose interest is unknown, and their application by means of an artificial neural network with unsupervised and competitive learning of the self-organising map or SOM (Self-Organising Map) type, and more specifically the Kohonen network, the underlying structure of these variables can be discovered and represented in a comprehensible way.

The obtained results are easily interpretable, allowing to recognize the grouping in the shape of patterns of architectural or urban objects represented, being allowed then a simple verification of the integration of a new building or object to the set of patterns that compose the network ordinance. These pattterns are called Ecotypes. The new objects accepted by this standard will form part of this standard, being integrated in the body of a new neural network, achieving this way a continued feedback of the ordinance. The proposed method is conformed as a decision-making aid method and as a assissance tool to the project of an ordinance. As a verification of the method, an implementation of an network ordinance is proposed in the historical center of Santa Fe in the province of Granada, comparing the obtained results with the in effecto planning represents.

error: Content is protected !!