Table 1.
Summary of related works.
Fig 1.
Proposed methodology for mental health classification in Thalassemia patients.
Table 2.
Detailed description of the dataset’s numerical features.
Fig 2.
3D t-SNE visualization of classification data: class 0 ("Good") vs. class 1 ("Bad").
Table 3.
Feature selection methods and selected features for the dataset.
Fig 3.
Feature importance plot based on AMSE-DFI for Thalassemia related mental health patients.
Table 4.
Model name and hyperparameter optimization.
Table 5.
Comprehensive performance comparison of ML models under different configurations.
Fig 4.
Performance comparison of ML and DL models without feature selection.
Fig 5.
Performance comparison of ML models using different feature selection methods.
Fig 6.
Performance comparison of ML models using different feature selection methods with SMOTE.
Fig 7.
(A) AUROC curve, (B) confusion matrix, and (C) log loss curve for the PSO-based LGBM model with SMOTE and AMSE-DFI feature selection.
Table 6.
Summary of statistical significance tests for top-performing models.
Fig 8.
Multi-dimensional analysis of mental health status across demographic, clinical, and socioeconomic factors — (A) Age and gender variation, (B) Diagnosis-based distribution, and (C) Economic class comparison.
Fig 9.
Multi-dimensional analysis of mental health status across healthcare and functional dimensions — (A) Medical expense distribution by mental health status, and (B) Physical functioning score comparison.
Fig 10.
LIME tabular explainer plot for predicting mental health status in Thalassemia patients.
Fig 11.
LIME feature importance plot for predicting mental health status in Thalassemia patients.
Table 7.
Comparison of methods and accuracy for Thalassemia prediction.