Disaggregating employment data to building level : a multi-objective optimisation approach

Abstract

The land use policies and development plans that are implemented in a city contribute to whether the city will be sustainable in the future. Therefore, when these policies are being established they should consider the potential impact on development. An analytical tool, such as land use change models, allow decision-makers to see the possible impact that these policies could have on development. Land use change models like UrbanSim make use of the relationship between households, buildings, and employment opportunities to model the decisions that people make on where to live and work. To be able to do this the model needs accurate data. When there is a more accurate location for the employment opportunities in an area, the decisions made by individuals can be better modelled and therefore the projected results are expected to be better. Previous research indicated that the methods that are traditionally used to disaggregate employment data to a lower level in UrbanSim projects are not applicable in the South African context. This is because the traditional methods require a detailed employment dataset for the disaggregation and this detailed employment dataset is not available in South Africa. The aim of this project was to develop a methodology for a metropolitan municipality in South Africa that could be used to disaggregate the employment data that is available at a higher level to a more detailed building level. To achieve this, the methodology consisted of two parts. The first part of the methodology was establishing a method that could be used to prepare a base dataset that is used for disaggregating the employment data. The second part of the methodology was using a multi-objective optimisation approach to allocate the number of employment opportunities within a municipality to building level. The algorithm was developed using the Distributed Evolutionary Algorithm in Python (DEAP) computational framework. DEAP is an open-source evolutionary algorithm framework that is developed in Python and enables users to rapidly create prototypes by allowing them to customise the algorithm to suit their needs The evaluation showed that it is possible to make use of multi-objective optimisation to disaggregate employment data to building level. The results indicate that the employment allocation algorithm was successful in disaggregating employment data from municipal level to building level. All evolutionary algorithms come with some degree of uncertainty as one of the main features of evolutionary algorithms is that they find the most optimal solution, and so there are other solutions available as well. Thus, the results of the algorithm also come with that same level of uncertainty. By enhancing the data used by land use change models, the performance of the overall model is improved. With this improved performance of the model, an improved view of the impact that land use policies could have on development can also be seen. This will allow decision-makers to draw the best possible conclusions and allow them the best possible opportunity to develop policies that will contribute to creating sustainable and lasting urban areas.