Depression is one of the most prevalent mental disorders in the world, which can worsen existing medical conditions and could lead to suicide if left untreated. . The goal of this project is to use features from facial, audio and GPS modalities that can be gathered from a smartphone to assess and track trajectories of depression. Features were extracted from the DAIC-WOZ and StudentLife datasets and the Naïve Bayes, random forest classifier, Support Vector Machine with stochastic gradient descent, and XGBoost classifiers were used to detect depression levels based on the PHQ score. The best performing classifier used was the XGBoost algorithm, with a mean accuracy of 0.82 for 2 bin classification and 0.639 for 3 bin classification. Features from the GPS modality had the highest metrics overall with a mean accuracy of 0.8875 for two 2-bin classification and 0.6625 for 3-bin classification.

• 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.

Creator

• Eteng, Praise

Publisher

• Worcester Polytechnic Institute

Identifier

• 57226
• E-project-033022-131721

Advisor

• Agu, Emmanuel O.

Year

• 2022

Date created

• 2022-03-30

Resource type

• Major Qualifying Project

Major

• Computer Science

Rights statement

• In Copyright