Article’s

Stress, anxiety, and depression have become increasingly prevalent among college students and working professionals, yet most affected individuals do not seek help until symptoms are already severe. Conventional screening methods—clinician-administered questionnaires and diagnostic interviews—are slow, costly, and inaccessible to many. This paper presents a web-based mental wellness assistant that uses machine learning and natural language processing to screen users for common mental health conditions and generate tailored self-management suggestions. Users provide input through a structured symptom questionnaire and a text-based chatbot; the responses are processed by an NLP pipeline and classified using Logistic Regression, Support Vector Machine, and Random Forest models. A severity score derived from the classification output determines the type of recommendation returned. In experimental evaluation, the Random Forest classifier achieved approximately 88% accuracy. The paper also identifies the system’s current limitations and the conditions under which broader deployment would be appropriate.