Stress level modeling and predictions are essential in recommending activities and interventions to individuals. While successful stress models have been proposed in the literature, there is still a missing connection between user engagement behaviors, interest in activities, and their stress levels. In this paper, we propose a novel multi-view tensor decomposition method for stress and user behavior modeling with heterogeneous data, which could provide personalized stress tracking and plausible user behavior modeling across time. To the best of our knowledge, it is the first method that could model user stress and behavior at the same time with multiple resources of data, such as stress measurement, activity rating, and engagement. Our experiments show that leveraging multiple resources of data could not only improve predictions with sparse data, but also results in discovering the underlying stress-activity patterns. We demonstrate the effectiveness of our proposed model on the dataset collected via a self-contained stress management mobile application.