LSTM-Based Longitudinal Prediction of Psychological Resilience: The Role of Self-Compassion, Meaning in Life, Cognitive Reappraisal, and Social Support

Objective: The present study aimed to investigate the longitudinal prediction of psychological resilience using an LSTM-based deep learning model by examining the dynamic contributions of self-compassion, meaning in life, cognitive reappraisal, and social support over time.Methods and Materials: This longitudinal study was conducted on ۴۲۸ participants recruited from Canada, with ۳۹۲ retained for final analysis across four time points over a ۱۲-month period. Data were collected using validated self-report instruments, including the Connor–Davidson Resilience Scale, Self-Compassion Scale, Meaning in Life Questionnaire, Emotion Regulation Questionnaire (cognitive reappraisal subscale), and the Multidimensional Scale of Perceived Social Support. Data preprocessing included multiple imputation and sequence structuring for time-series analysis. The primary analytical approach involved the application of Long Short-Term Memory (LSTM) neural networks implemented in Python using TensorFlow and Keras. The dataset was divided into training, validation, and test sets (۷۰/۱۵/۱۵), and model performance was evaluated using Mean Squared Error (MSE), Root Mean Squared Error (RMSE), and R². SHAP analysis was conducted to determine the temporal importance of predictors.Findings: The LSTM model demonstrated strong predictive performance (R² = ۰.۶۶, RMSE = ۲.۵۲), indicating substantial explained variance in psychological resilience. Self-compassion emerged as the most significant predictor (β ≈ .۵۴, p < .۰۰۱), followed by social support (β ≈ .۵۱, p < .۰۰۱), meaning in life (β ≈ .۴۹, p < .۰۰۱), and cognitive reappraisal (β ≈ .۴۳, p < .۰۰۱). Longitudinal analyses revealed significant increases in resilience and all predictor variables across time (p < .۰۱). SHAP results indicated that self-compassion and meaning in life showed increasing contributions over time, whereas social support demonstrated stronger early influence and cognitive reappraisal maintained a stable effect across all time points.Conclusion: The findings highlight the dynamic and multifactorial nature of psychological resilience, emphasizing the central role of self-compassion and the evolving contributions of internal and external resources over time. The integration of LSTM modeling with explainable AI provides a robust framework for capturing temporal patterns and enhancing predictive accuracy in psychological research. Objective: The present study aimed to investigate the longitudinal prediction of psychological resilience using an LSTM-based deep learning model by examining the dynamic contributions of self-compassion, meaning in life, cognitive reappraisal, and social support over time. Methods and Materials: This longitudinal study was conducted on ۴۲۸ participants recruited from Canada, with ۳۹۲ retained for final analysis across four time points over a ۱۲-month period. Data were collected using validated self-report instruments, including the Connor–Davidson Resilience Scale, Self-Compassion Scale, Meaning in Life Questionnaire, Emotion Regulation Questionnaire (cognitive reappraisal subscale), and the Multidimensional Scale of Perceived Social Support. Data preprocessing included multiple imputation and sequence structuring for time-series analysis. The primary analytical approach involved the application of Long Short-Term Memory (LSTM) neural networks implemented in Python using TensorFlow and Keras. The dataset was divided into training, validation, and test sets (۷۰/۱۵/۱۵), and model performance was evaluated using Mean Squared Error (MSE), Root Mean Squared Error (RMSE), and R². SHAP analysis was conducted to determine the temporal importance of predictors. Findings: The LSTM model demonstrated strong predictive performance (R² = ۰.۶۶, RMSE = ۲.۵۲), indicating substantial explained variance in psychological resilience. Self-compassion emerged as the most significant predictor (β ≈ .۵۴, p < .۰۰۱), followed by social support (β ≈ .۵۱, p < .۰۰۱), meaning in life (β ≈ .۴۹, p < .۰۰۱), and cognitive reappraisal (β ≈ .۴۳, p < .۰۰۱). Longitudinal analyses revealed significant increases in resilience and all predictor variables across time (p < .۰۱). SHAP results indicated that self-compassion and meaning in life showed increasing contributions over time, whereas social support demonstrated stronger early influence and cognitive reappraisal maintained a stable effect across all time points. Conclusion: The findings highlight the dynamic and multifactorial nature of psychological resilience, emphasizing the central role of self-compassion and the evolving contributions of internal and external resources over time. The integration of LSTM modeling with explainable AI provides a robust framework for capturing temporal patterns and enhancing predictive accuracy in psychological research.

کلیدواژه ها:

psychological resilience ، self-compassion ، meaning in life ، cognitive reappraisal ، social support