KMAN Publication Inc. (KMANPUB) Journal of Adolescent and Youth Psychological Studies (JAYPS) 2981-2526 7 Serial Number 44 2026 04 10 Predicting Self-Esteem Instability Using Rejection Sensitivity, Daily Stress Reactivity, Social Feedback Valence, and Emotional Reactivity in LSTM Models Predicting Self-Esteem Instability Using Rejection Sensitivity, Daily Stress Reactivity, Social Feedback Valence, and Emotional Reactivity in LSTM Models 10.61838/kman.jayps.5268 EN Journal Article 2025 10 24 <table> <tbody> <tr> <td> <p><strong>Objective:</strong> To evaluate the efficacy of a Long Short-Term Memory ( ) neural network in predicting intra-individual self-esteem instability utilizing continuous ecological momentary assessment data encompassing trait rejection sensitivity, daily stress reactivity, social feedback valence, and momentary emotional reactivity.</p> <p><strong>Methods and Materials:</strong> A longitudinal Ecological Momentary Assessment study was conducted over days with a final sample of young adults from Spain. Participants completed the Rejection Sensitivity Questionnaire and responded to multiple daily prompts assessing momentary stress, social feedback valence, emotional reactivity, and state self-esteem. An neural network was designed to capture the non-linear, time-dependent relationships within this multi-dimensional data, and its performance was benchmarked against traditional linear regression and standard Recurrent Neural Network ( ) architectures.</p> <p><strong>Findings:</strong> The proposed architecture significantly outperformed baseline models, capturing of the variance in self-esteem instability ( ) with a low prediction error ( ). Permutation feature importance analysis revealed that momentary emotional reactivity and trait rejection sensitivity were the strongest predictors of self-esteem fluctuations. Additionally, an ablation study on look-back windows identified a -hour period ( &nbsp;assessments) as the optimal temporal framework for forecasting instability.</p> <p><strong>Conclusion:</strong> Advanced deep learning models, specifically networks, can accurately decode the complex, non-linear temporal dynamics of self-esteem, providing a powerful computational foundation for future personalized, real-time psychological interventions.</p> </td> </tr> </tbody> </table> https://www.journals.kmanpub.com/index.php/jayps/article/download/5268/9574