DS:MQP:Deep Learning for Mental Health Screening using Smartphone Data Public
Downloadable Contentopen in viewer
Almost 300 million people across the world suffer from depression. This is a widespread issue in which a majority of those suffering do not receive treatment. Screening for depression using smartphone data could result in less biased, more proactive results and could be accomplished by recognizing text patterns in people’s messages. To explore this potential, we analyzed previously collected text message data through the application of lexical category featurization and a deep learning model, BERT. While we experimented with existing data, we also researched methods for further data collection. We concluded that the Bag of Words method of categorizing words outperformed the Empath method. We also determined that the methods of lexical category featurization produced higher accuracies than the BERT models, and more advanced BERT models were prone to overfitting. By visualizing the results, we discovered that those with depression discussed less about personal topics than asymptomatic people.
- This report represents the work of one or more WPI undergraduate students submitted to the faculty as evidence of completion of a degree requirement. WPI routinely publishes these reports on its website without editorial or peer review.
- UN Sustainable Development Goals
- Date created
- Resource type
- Rights statement
Permanent link to this page: https://digital.wpi.edu/show/gb19f908t