Development of a Dempster-Shafer Model for Diagnosing Mental Health Disorders on Social Media

The rapid proliferation of social media has been paralleled by an increase in mental health concerns, often reflected in users' online behaviors and text posts. Traditional methods for diagnosing mental health disorders do not effectively leverage this vast amount of data, nor do they address the ambiguities and uncertainties inherent in natural language processing of social media content. A novel diagnostic model based on the Dempster-Shafer evidence theory is developed specifically for analyzing social media text. The model aims to enhance the precision of mental health disorder diagnoses by incorporating probabilistic reasoning and handling uncertainty in the data more effectively than traditional binary classification methods. It integrates linguistic cues, sentiment analysis, and behavioral indicators extracted from user posts to assess the likelihood of various mental health conditions. We collected a dataset comprising social media posts tagged with mental health keywords, verified by clinical annotations. The Dempster-Shafer model was trained with features derived from text analytics, including sentiment scores, thematic elements, and engagement metrics. The model’s performance was evaluated against standard classification models through metrics such as accuracy, sensitivity, and specificity. The Dempster-Shafer model demonstrated superior capability in handling the uncertainty and imprecision of social media data, showing an improvement in diagnostic accuracy by 15% over traditional models. The findings suggest that applying Dempster-Shafer theory in mental health diagnostics can effectively utilize the nuanced and sparse data available on social media, offering a promising tool for early identification and intervention in mental health issues.