Enhanced Interpretability in Depression Classification with the Use of Machine Learning Models with Explainable AI

Depression, a prevalent mental health concern, demands efficient and accurate classification methods to support timely interventions. This study utilizes two mental health datasets to classify depression levels into high, medium, and low categories. Data preprocessing and visualization techniques were applied to extract meaningful patterns, followed by the implementation of six widely used machine learning algorithms: Random Forest (RF), Logistic Regression (LR), Support Vector Machine (SVM), Decision Tree (DT), K-Nearest Neighbors (KNN), and Naive Bayes (NB). Among these, the RF model demonstrated the highest accuracy, achieving 0.98 without cross-validation and 0.99 with cross-validation. Class-wise evaluation using precision, recall, and F1 score further established RF’s superior performance. To enhance interpretability and transparency, Local Interpretable Model-agnostic Explanations (LIME) was employed, providing insights into the RF model’s decision-making process. This research underscores the efficacy of the RF model for depression classification and highlights the importance of integrating explainable AI techniques in mental health diagnostics.