Detecting Symptoms of Depression in Social Media Data using Machine Learning

Author

Eunkyu Park

Date of Award

2020

First Advisor

Amanda Landi

Second Advisor

Michael Bergman

Abstract

Based on the sentence, "you are what you post", there have been continuous studies conducted to identify depression by leveraging social media data. Depression detection traditionally is done through a face-to-face meeting with a clinical professional. However, only 52% of patients come in seeking help early enough for effective intervention (Boerema AM, 2016). Factors such as lack of knowledge, ignorance, prejudice and social stigma against people with mental illness and expectation of discrimination of people who have mental illness cause these delays in treatment (Henderson, 2013). Nowadays, social media is a place in which people post their feelings more often than talking to a clinician about their emotions. Therefore, much research has been conducted to detect depressive symptoms in social media data. This research completes three different tasks with some novel approaches. First, we apply word cloud analysis to understand different word usages between depressed and nondepressed users. Second, we train a classification model for detecting depressed users by utilizing a different python-library called Fasttext. In addition, implementing topic modeling to yield insight into how depressed and non-depressed individuals use language differently on social media.

Recommended Citation

Park, Eunkyu, "Detecting Symptoms of Depression in Social Media Data using Machine Learning" (2020). Senior Theses. 1435.
https://digitalcommons.bard.edu/sr-theses/1435