Benjumeda Wynhoven Isabel María, Lepe Claudio Córdova
Facultad de Artes Liberales, Universidad Adolfo Ibáñez. Campus Viña del Mar, Valparaíso, Chile.
Interdisciplinary Center for Biomedical Research and Health Engineering. University of Valparaiso, Valparaíso, Chile.
PLoS One. 2025 Jun 3;20(6):e0323945. doi: 10.1371/journal.pone.0323945. eCollection 2025.
Intimate partner violence (IPV) is a serious social problem in Chile. Understanding the patterns of internalization and the motivations maintaining it is crucial to design optimal treatments that ensure adherence and completeness. This, in addition, is essential to prevent revictimization and improve the quality of life of both victims and their children.The present study analyzes the success of a psychological treatment offered by a Chilean foundation helping IPV victims. A database analysis containing 1,279 cases was performed applying classical statistics and artificial intelligence methods. The aim of the research was to search for cluster grouping and to create a classification model that is able to predict IPV treatment completeness. The main results demonstrate the presence of two main clusters, one including victims who completed the treatment (cluster 1) and a second one containing victims who did not complete the treatment (cluster 2). Cluster classification using an XGBoost model of the treatment completeness had an accuracy of 81%. The results showed that living with the aggressor, age and educational level had the greatest impact on the classification. Considering these factors as input variables allow for a higher precision on the treatment completeness prediction. To our knowledge, this is the first study performed in Chile that uses AI for cluster grouping and for analyzing the variables contributing to the success of an IPV victims' treatment.
亲密伴侣暴力（IPV）在智利是一个严重的社会问题。了解内化模式以及维持这种模式的动机对于设计确保依从性和完整性的最佳治疗方法至关重要。此外，这对于防止再次受害以及改善受害者及其子女的生活质量也至关重要。本研究分析了智利一个帮助亲密伴侣暴力受害者的基金会提供的心理治疗的成效。运用经典统计学和人工智能方法对包含1279个案例的数据库进行了分析。该研究的目的是寻找聚类分组，并创建一个能够预测亲密伴侣暴力治疗完整性的分类模型。主要结果表明存在两个主要聚类，一个聚类包括完成治疗的受害者（聚类1），另一个聚类包含未完成治疗的受害者（聚类2）。使用XGBoost模型对治疗完整性进行聚类分类的准确率为81%。结果显示，与施暴者同居、年龄和教育水平对分类的影响最大。将这些因素作为输入变量可提高治疗完整性预测的精度。据我们所知，这是智利首次使用人工智能进行聚类分组并分析有助于亲密伴侣暴力受害者治疗成功的变量的研究。
