Identification and importance analysis of osteoporosis risk factors in perimenopausal women based on the Transformer-LSTM hybrid model and SMOTE

This study aims to identify key risk factors for osteoporosis in perimenopausal women using a novel analytical framework that combines a Transformer-LSTM hybrid model with the SMOTE algorithm for data balancing. The findings highlight total type 1 collagen N-terminal propeptide (TP1NP) as the most critical physiological indicator, along with other significant factors such as educational level and body weight. The proposed model demonstrates improved classification performance, suggesting its potential for enhancing early diagnosis and personalized interventions for osteoporosis in this population.