Social Media Insight: Predicting Self-Harm Trends

Poisoning or injuring oneself can cause death or injury without intent. This is self-harm. Self-harm harms the perpetrators and the nation's economy. Studies have linked increased urbanization in emerging nations to rising self-harm and new technology. National self-harm patterns may be important for policymakers and public health experts to predict. This would allow them to immediately fix issues or avoid disasters. Studies have forecasted population-level self-harm patterns in many nations using simple statistics. In certain nations, prior data may be scarce or insufficient to make reliable projections. This makes it tougher to FAST assess and anticipate national self-harm. This essay proposes FAST, a technique that uses mental indicators from social media to forecast national self-harm trends. These markers can represent community mental health and help forecast self-harm patterns. Language-agnostic algorithms are taught to first identify mental indicators in gathered social media communications. These signals form multivariate time series. The time-delay embedding method then places these events in time. Last, many machine learning regressors are tried for future prediction. Tweets with 12 mental indicators may predict self-harm-related fatalities and injuries, according to a Thailand research. The suggested method predicted self-harm deaths and injuries. Better than ARIMA baseline by 43.56% and 36.48%. We think our research is the first to predict national self-harm trends using social media data. The findings improve self-harm prediction and establish the framework for new social network-based apps that forecast socioeconomic aspects. We used the top machine learning algorithms, Decision Tree and Voting Regressor. We found less MAE mistakes with these techniques.