ANALYZING EMOTIONAL PATTERNS IN SOCIAL MEDIA FOR MENTAL HEALTH DISORDER DETECTION

Mental health disorders including depression and anorexia still afflict millions of people globally; their symptoms are often overlooked as they are so subtle and distinctive. Early detection—which is still rather challenging—is what makes effective intervention possible. Social media channels give a rare opportunity to see people's emotional reactions in real time via their written works. This work investigates the emotional patterns in user posts using Natural Language Processing (NLP) methods in search of potential markers of mental health problems. We propose a computational framework including sentiment polarity, emotional categories, and language cues to detect emotional aspects in user-generated social media content. Then trained using these traits a Decision Tree classifier selected for its simplicity of use, interpretability, and feature-based decision making capabilities. Decision trees help to successfully distinguish between impacted and non-affected persons by use of emotional-based attributes. Two public ally datasets related to depression and anorexia are used for testing the algorithm. Thanks to the interpretable structure of decision trees, our approach provides better transparency; it also achieves competitive performance quite similar to state-of-the-art models. Apart from raising detection accuracy, the coupling of decision tree classification with NLP-driven emotional analysis generates chances for explainable artificial intelligence in applications related to mental health. This method prepares the stage for real-time, non-invasive support systems emphasizing early mental health care detection and intervention.