Predicting national suicide numbers with social media data

Area of Science:

• Social Sciences
• Public Health
• Computational Social Science

Background:

• Suicide is a significant public health issue influenced by social and environmental factors.
• South Korea has a high suicide rate and extensive social media usage.
• Traditional predictors of suicide include social, economic, and meteorological variables.

Purpose of the Study:

• To investigate the potential of social media data for predicting population-level suicide rates.
• To compare the predictive power of social media variables against traditional predictors.
• To develop and validate a suicide prediction model using social media data.

Main Methods:

• Utilized weblog entries related to suicide and dysphoria as social media variables.
• Incorporated classical social, economic, and meteorological data.
• Developed a prediction model using 2-year training data (2008-2009) and validated it on subsequent year's data (2010).
• Performed sensitivity analysis controlling for celebrity suicide events.

Main Results:

• Both social media variables (suicide-related and dysphoria-related) were strongly associated with suicide frequency.
• Dysphoria-related weblogs reflected longer secular trends, while suicide-related entries showed higher variability and reactivity to events.
• Social media variables, particularly dysphoria, outperformed economic indicators like consumer price index and unemployment rate in the final model.
• The prediction model demonstrated robustness during validation and sensitivity analysis.

Conclusions:

• Social media data offers a valuable, novel resource for national suicide forecasting.
• The findings suggest that social media analytics can enhance suicide prevention strategies.
• Digital footprints, specifically concerning mental state and suicidal ideation, provide insights into population-level mood and affect.