KMAN Publication Inc. (KMANPUB) Applied Family Therapy Journal (AFTJ) 3041-8798 7 Serial Number 32 2026 03 01 Using Deep Learning to Classify Family Functioning Profiles Based on Parental Attachment, Emotion Socialization, and Child Behavioral Outcomes Using Deep Learning to Classify Family Functioning Profiles Based on Parental Attachment, Emotion Socialization, and Child Behavioral Outcomes 1 11 10.61838/kman.aftj.5215 EN Journal Article 2026 01 12 <p><strong>Objective:</strong> The objective of this study was to identify distinct family functioning typologies based on parental attachment, emotion socialization practices, and child behavioral outcomes, and to subsequently develop a deep learning model for the accurate automated classification of these profiles.</p> <p><strong>Methods and Materials:</strong> This study utilized a cross-sectional design involving two-parent families from Greece. Data were collected using secure online self-report questionnaires, including the Experiences in Close Relationships-Revised for parental attachment, the Coping with Children’s Negative Emotions Scale for parental emotion socialization, and the Strengths and Difficulties Questionnaire for child behavioral outcomes. Data analysis proceeded in two phases: first, Latent Profile Analysis (LPA) was utilized to identify unobserved family profiles. Second, a Multi-Layer Perceptron (MLP) deep learning model was developed to classify families into these established profiles. The dataset was partitioned into training ( ), validation ( ), and testing ( ) segments to evaluate the model’s predictive performance.</p> <p><strong>Findings:</strong> Latent Profile Analysis successfully identified three distinct family functioning profiles: Adaptive Family Functioning ( , ), Anxious-Inconsistent Functioning ( , ), and High-Risk/Avoidant-Punitive Functioning ( , ). The trained Multi-Layer Perceptron deep learning model demonstrated exceptional classification capabilities, achieving an overall accuracy of on the unseen test set. The model yielded high F1-scores for each respective class: for the Adaptive profile, for the Anxious-Inconsistent profile, and for the High-Risk profile, with the few misclassifications occurring primarily between the Anxious-Inconsistent and High-Risk groups.</p> <p><strong>Conclusion:</strong> The robust identification of multidimensional family profiles and their highly accurate classification via an artificial neural network provides a powerful, innovative methodological framework to enhance targeted clinical assessments and tailor psychological interventions.</p> https://www.journals.kmanpub.com/index.php/aftj/article/download/5215/9474