A GUI-Based tool for social anxiety assessment: A comparative study of machine learning models

This paper presents a Graphical User Interface (GUI)-based tool for assessing social anxiety severity using machine learning models. The system is designed to assist individuals who feel uncomfortable engaging with real-life therapists by offering an AI-driven, user-friendly alternative. A structured questionnaire was administered to 500 Jordanian participants, and responses were used to train and evaluate three classification models: Artificial Neural Networks (ANN), Decision Trees (DT), and Random Forests (RF). The dataset was split into 400 training and 100 testing samples. ANN achieved a validation accuracy of 96.5% and test accuracy of 87%. DT reached 98.5% validation and 98% test accuracy. RF outperformed both, achieving 100% accuracy on validation and test sets. These results demonstrate the feasibility of machine learning models in accurately classifying social anxiety levels based on self-reported data. The tool was implemented in MATLAB with an interactive GUI to enhance accessibility and usability. By encouraging users to engage with the assessment process digitally, the system offers a supportive environment for individuals hesitant to seek traditional therapy. This work contributes to the development of AI-assisted mental health tools that are both effective and sensitive to user experience.