In this paper, we propose an individual income level inference method based on travel behavior of urban residents. Firstly, we define 13 residents' travel behavior features from 3 different aspects. These features are the mobility indicators based on trajectories, home-based travel characteristics, and activity chain patterns. Secondly, 80% of the daily travel diary data of residents in Guangzhou in 2013 is taken as the training sample and the remaining as the test data. Thirdly, the random forest method is trained and used to infer individual's income level. The experiment results show that our proposed method can obtain an overall accuracy of 80%. Among 13 travel behavior features, the home-based features (i.e. the mode of the travel distance from home during working time (9:00 AM to 18:00 PM)), the activity chain pattern, and travel distance and scope related features (i.e., the maximum distance between two successive activity points, and radius of gyration) have high feature importance. However, the features representing the spatial heterogeneity of the activity points have relatively low importance, such as the spatial diversity.