404 / 2019-02-26 22:47:52

Data for Informal Cities: Land Use Classification Using Twitter Data in Nairobi

random forest,informal city,geo-tweets,land use classficiation

全体主题 > Big data in Urban Planning and Management

Land-use and land-cover classification is essential for urban planning and management. However, the challenge of shifting socioeconomic informality, rapid urban growth, and incapacity of information management in cities of developing countries make the relevant data rarely available. There are established techniques using satellite images as a solution for landcover classification by physical characteristics (e.g. texture). Yet social information (e.g. population density, human activity) that mostly defines land uses was not captured in remote sensing data, which critically limit its performance in mapping urban areas. Our research explores a way of enhancing urban mapping using emerging human mobility data. In particular, this work applies a random forest method on geo-tweets for land use classification in Nairobi. The results gave fine predictions in urban built-up areas with various types of settlements (e.g. compact high-rise, open low-rise) identified. A discussion on the potential and limitations of geo-tweets was presented based on our analysis results.