Introduction In recent years, ubiquitous devices have penetrated people's lives, and numerous studies have been conducted to find behavioral and emotional patterns affecting health and well-being [1],[2]. Especially in mental healthcare, till now, the tracking of the patient's conditions relied solely on doctor appointments and self-reported surveys, which are timeconsuming and might lack objectivity. During their university years, students often suffer from accumulated stress. Thus, early diagnoses and improved monitoring are becoming vital. Exploiting the StudentLife dataset [3], a structured approach to predict the self-reported PANAS [4] Negative Affect (NA), consult students and reduce university drop-outs is briefly introduced. Data Sources The StudentLife sensing app captured the daily impact of assignments on the activity, mood, sociability, well-being, and academic performance of 48 students throughout the semester. The StudentLife dataset contains sensor data, Ecological Momentary Assessment (EMA) [5], survey responses, and educational data. The first results show significant correlations between smartphone objective sensor data and the student body's mental health and academic profiles. The PANAS questionnaire is a frequently used instrument assessing positive and negative affect and, in our work, serves as the ground truth. In this work, we explore the data types presented in Table 1 and investigate if we can predict NA from the objective sensing data. Additionally, there are data depicting academic performance, reporting the Grade Point Averages (GPAs), and the usage of the student forum called Piazza.
Random Forest (RF) regressor best predicts PANAS NA. We plan to move to a Deep Learning Architecture as a next step. Additionally, we intend to study specific social groups separately, e.g., males and females, first- and second-gen university students, capturing personalized contexts of college students.
This project has received funding from the European Union’s Horizon 2020 research and innovation programme under the Marie Skl odowska Curie grant agreement No 813162. The content of this paper reflects only the authors’ view and the Agency and the Commission are not responsible for any use that may be made of the information it contains.