KMAN Publication Inc. (KMANPUB) Applied Family Therapy Journal (AFTJ) 3041-8798 7 Serial Number 32 2026 03 01 Unsupervised Clustering of Intimate Partner Violence Typologies: Analyzing Communication Deviance, Attachment Anxiety, and Trait Impulsivity via Machine Learning Unsupervised Clustering of Intimate Partner Violence Typologies: Analyzing Communication Deviance, Attachment Anxiety, and Trait Impulsivity via Machine Learning 1 11 10.61838/kman.aftj.5228 EN Journal Article 2026 01 09 <p><strong>Objective:</strong> This study aimed to identify distinct typologies of intimate partner violence (IPV) risk by analyzing the interplay of communication deviance, attachment anxiety, and trait impulsivity using unsupervised machine learning within a community sample.</p> <p><strong>Methods and Materials:</strong> A cross-sectional quantitative design was utilized with a sample of individuals from heterosexual couples in Poland. Data were collected via online self-report measures, including the Revised Conflict Tactics Scales (CTS2), the Communication Deviance Scale, the Experiences in Close Relationships-Revised (ECR-R) for attachment anxiety, and the Barratt Impulsiveness Scale (BIS-11). A k-means clustering algorithm was applied to the standardized psychological variables to derive typologies, which were subsequently profiled and validated against IPV severity using one-way ANOVAs.</p> <p><strong>Findings:</strong> The k-means algorithm yielded a robust three-cluster solution. Cluster 1, the “Regulated/Secure” group ( , ), exhibited low scores across all three risk factors and reported minimal IPV. Cluster 2, the “Anxious-Impulsive” group ( , ), displayed elevated attachment anxiety and trait impulsivity alongside moderate communication deviance, corresponding to significantly higher rates of psychological and physical aggression. Cluster 3, the “Dysregulated/Deviant” group ( , ), demonstrated the highest scores across all three psychological constructs and reported the most severe, multifaceted IPV, including significantly elevated levels of sexual coercion and physical injury.</p> <p><strong>Conclusion:</strong> The identification of these distinct psychological risk profiles highlights the fundamental heterogeneity of IPV perpetration and underscores the critical necessity for tailored, person-centered screening and intervention strategies.</p> https://www.journals.kmanpub.com/index.php/aftj/article/download/5228/9500