Monintoring Student Depression by Mobile Devices with Autoencoder Networks

Abstract

The timely assessment of mental health is difficult since we lack the objective measurements of symptoms, especially for the Covid-19 pandemic quarantined students. Fortunately, smart phones can capture the real-world data such as the GPS traces and the phone active time et.al that link the behavioral patterns to the mental health. However, recent studies are based on a very small size of participants and only collect fewer phone features, which means that the effective predicting models which require various features are hardly adopted. In this paper, we develop an android application to record multidimensional data of users as well as a PHQ-9 and a SAS questionary, and we distribute it to 176 college students to collect larger scale data when in quarantine period. To address the unprecise problem of handcrafted feature extraction, we design an autoencoder machine learning model to monitor the student mental health. Extensive experiments indicate that the performance of the proposed method improves its F-1 score for PHQ-9 and SAS by 5% and 6% to the state of the current studies, respectively.

Cite this work

No public copy is stored for this record, so only the metadata is cited.

Stored only when both papers are in the corpus — 1 citations exist upstream.

Papers that cite the same references — the relational fallback while embeddings are still backfilling.