This study developed a new approach to link residents' daily travel behavior and their socio-economic attributes by inferring residents' income level from their activity space, activity sequence, and geographic exposure. A classification and regression tree (CART) analysis on data from Guangzhou, China provided three key outcomes. First, residents' income level can be inferred with considerable accuracy through their daily activity and geographic exposure. Second, geographic exposure and activity sequence variables are useful in the classification process, with vital information possibly overlooked if only mobility features are used. Third, people in different income groups can be identified by different characteristics, with high-income earners identified by their preference for particular socio-economic environments, low-income earners identified by their fixed lifestyles and dependence on affordable and convenient environments, and middle-income earners identified by more comprehensive characteristics. This study has improved our understanding of social group diversity from a spatio-temporal behavior perspective. It's also instructive in merging the attributes of conventional data with those of anonymized big data. This benefits future big data research and the development of policies supported by big data. • Residents' income level can be inferred with considerable accuracy through their daily activity and geographic exposure. • Including only mobility features in inferring residents' socio-economic attributes possibly overlooked vital information. • People in different income groups can be identified by different daily behavior characteristics • Through the linkage from residents' daily behavior to their socio-economic attributes, information of big data and conventional data can be merged. [ABSTRACT FROM AUTHOR]