Big data analytics on social networks for real-time depression detection

Area of Science:

• Computational social science
• Mental health informatics
• Machine learning applications

Background:

• Increased depression cases during the COVID-19 pandemic necessitate novel detection methods.
• Social media platforms like Twitter offer a rich source of user sentiment data.
• Real-time depression detection can aid in timely intervention and support.

Purpose of the Study:

• To propose and evaluate a big data analytics model for real-time depression detection using Twitter data.
• To analyze demographic characteristics and user opinions for depression identification.
• To compare the efficacy of various machine learning techniques for this task.

Main Methods:

• Collected Twitter user data over two months, including demographic information and posts.
• Utilized the Patient Health Questionnaire-9 (PHQ-9) as an outcome measure for depression.
• Applied and compared five machine learning techniques: Support Vector Machine, Decision Tree, Naïve Bayes, Random Forest, and Deep Learning.

Main Results:

• The Random Forest machine learning technique demonstrated superior accuracy in detecting depression compared to other methods.
• The developed model effectively captured depressive moods by analyzing user demographics and text sentiment.
• Analysis of demographic characteristics and opinions proved valuable for depression detection.

Conclusions:

• A novel model combining demographic data and Twitter sentiment analysis can effectively detect depression.
• Machine learning, particularly Random Forest, shows promise for real-time mental health monitoring on social media.
• This approach represents a significant step towards reducing depression-induced suicide rates.