Ensemble-Based Machine Learning Models for Classifying Intervention-Related Psychological Outcome Patterns in Menopausal Women Treated With Pimpinella anisum.

Abstract

Background: Menopause, a natural biological transition in the life of women, is often associated with significant psychological disturbances such as depression, anxiety, and stress. However, early identification and management of psychological disorders in menopausal women remain under-addressed due to limited specialist access and a lack of systematic screening approaches.

Purpose: This study aimed to develop ensemble-based machine learning (ML) models to classify intervention-related clinical and psychological outcome patterns in menopausal women using clinical and questionnaire-based data collected during a controlled intervention study.

Method: Psychological and menopausal symptoms were assessed using standardized tools, including the Depression Anxiety Stress Scale-21 (DASS-21), modified Kupperman index (MKI), vaginal health index (VHI), and menopause symptoms treatment satisfaction questionnaire (MS-TSQ). Statistical analyzes were performed using chi-square tests and independent t-tests to evaluate group differences. Three ensemble classifiers, random forest (RF), gradient boosting (GB), and AdaBoost (AB), were evaluated using repeated cross-validation.

Result: Among these, the RF classifier achieved the best performance with a sensitivity of 93.3%, a specificity of 90.0%, an accuracy of 91.7%, a precision of 90.3%, and an F1-score of 91.8%.

Conclusion: These results indicate that ensemble ML models, particularly RF, can serve as effective supplementary tools for psychological disorder risk stratification in menopausal care, aiding clinicians in decision-making. However, these models are intended to complement, not replace, clinical assessments.