This paper investigates the activity behavior of residents of Paraisópolis, the second largest slum of São Paulo (Brazil). The study used data from a survey and one week of GPS traces of a sample of residents. Location data was the basis to infer individual stays and points of interest. Stays were clustered into 6 classes, based on spatial, temporal, repetition and sequence variables characterizing each stay. These stays classes were used to describe individual weekly activity patterns. Individuals were then clustered into 7 categories, based on the similarity of their activity patterns, as described by measures of intensity, variation and repetition. Finally, each group was analyzed in terms of its demographic and socioeconomic composition. Results reveal considerable coherence, confirming expected relationships between the weekly activity patterns and individuals’ attributes. It should be highlighted that more than half of sampled residents were classified into groups with diversified behavior. This result, considering the high density and mixed land use of the Paraisópolis area, reinforces the idea that modelling efforts, even in poorer areas, need to consider activity patterns beyond the more usual simple commute. The article also demonstrates how new multiday data collection methods can contribute to improving the access to hard-to-reach groups, like slums residents.

本文调查了圣保罗（巴西）第二大贫民窟Paraisópolis居民的活动行为。该研究使用了一项调查的数据以及一个居民样本一周的GPS跟踪数据。位置数据是推断个人住宿和兴趣点的基础。根据每次住宿的特征，时空，重复和顺序变量，将其分为6类。这些停留类别用于描述个人每周的活动方式。然后根据强度，变异和重复的度量，根据活动模式的相似性，将个人分为7类。最后，根据人口统计和社会经济构成对每个群体进行了分析。结果显示出相当的连贯性，确认每周活动模式和个人属性之间的预期关系。应该强调的是，超过一半的抽样居民被分为行为多样化的群体。考虑到Paraisópolis地区的高密度和混合土地利用，这一结果加强了这样的思想，即即使在较贫困的地区，建模工作也需要考虑更普通的通勤方式以外的活动模式。本文还演示了新的多日数据收集方法如何有助于改善对贫民窟居民等难以到达的群体的访问。强化了这样一种观念，即即使在较贫困的地区，建模工作也需要考虑超出通常的简单通勤之外的活动模式。本文还演示了新的多日数据收集方法如何有助于改善对贫民窟居民等难以到达的群体的访问。强化了这样一种观念，即即使在较贫困的地区，建模工作也需要考虑超出通常的简单通勤之外的活动模式。本文还演示了新的多日数据收集方法如何有助于改善对贫民窟居民等难以到达的群体的访问。