This pilot study explores the application of Artificial Intelligence (AI) in the diagnosis and treatment of tinnitus, a prevalent auditory disorder, by leveraging advanced machine learning algorithms. Utilizing the Neurofeedback Tinnitus Dataset, the study focuses on developing and optimizing a multi-label classification model for accurate tinnitus diagnosis. Initially, a Random Forest classifier is employed to identify tinnitus-related patterns within the data. To enhance diagnostic accuracy and predictive performance, the study integrates advanced algorithms including LightGBM, Long Short-Term Memory (LSTM) networks, AdaBoost, and XGBoost. These models are selected for their robustness in handling high-dimensional, complex datasets and their proven efficacy in classification tasks. Furthermore, the implementation of SHAP (SHapley Additive exPlanations) provides model interpretability, enabling clinicians to comprehend the underlying rationale behind AI-driven predictions. This explainability fosters greater trust and transparency in clinical decision-making. Overall, the research contributes to the advancement of AI-assisted auditory healthcare, offering a foundation for personalized tinnitus treatment strategies and improved management practices through interpretable machine learning.

Keywords: Artificial Intelligence · Tinnitus Diagnosis · Machine Learning · Random Forest · LightGBM · LSTM · AdaBoost · XGBoost · SHAP · Explainability · Multi-label Classification.