Uma abordagem orientada a dados para a criação de um indicador de habitabilidade baseado na API da UBER

Abstract

One of the global dilemmas in the last decades has been the accelerated urban transition. Therefore, promoting sustainable urban development to accommodate population growth is extremely important. Under those circumstances, the concept of liveability arises, defined as a principle that combines economic, social and environmental attributes to promote quality of life and human well-being, and it has been widely discussed in the New Urban Agenda (NUA) adopted by the United Nations since 2016. NUA defines policies to promote the consolidation of Sustainable Development Goals (SDGs), with its Goal 11 focusing on a pro-urban future. To achieve those goals is recommended the use of indicators to supervise their implementation, thus the liveability concept can be associated with an indicator for that purpose. However, there are known issues related to data unavailability, poor quality and aggregation, that make the SDGs monitoring difficult. Considering the described scenario, this work proposes a liveability indicator that combines traditional data from census and other official surveys with alternative data sources, such as data from Uber, a popular ride-sharing service. Assuming that Uber service behavior can act as a proxy to liveability, a data science approach based on exploratory and spatial data analysis was conducted using Uber’s Estimated Time of Arrival (ETA) data sourced for the Brazilian city of Natal (RN). A data acquisition structure was developed and an investigation of missingness mechanism was performed in order to deal with missing data by means of a multiple imputation technique. This data science approach aims to build a composite indicator which can portray at some level the liveability for that city. In general, to create the proposed indicator, it was performed a multivariate analysis followed by the execution of weighting and aggregation methods on Uber and traditional surveys data. The proposed methodology was applied at two different spatial aggregation levels: Neighborhoods and Human Development Unities (HDUs). Results showed how the Uber service oscillates spatially and how it reacts to weather variations, festivals, and other events, as well as its relations with existing social and infrastructural indicators. It was also observed that different spatial aggregation levels affect the Uber ETA and its relations with socioeconomic variables. Finally, the proposed indicator was created at HDU scale to be applied in sustainable development monitoring. Furthermore, it was concluded that West and North administrative regions of Natal predominantly have localities with the worst liveability indicators.