Support Vector Machine Classification of High- and Low-Functioning Families Based on Communication, Adaptability, Emotional Security, and Conflict Management

Abstract

Objective: The present study aimed to develop and evaluate a Support Vector Machine (SVM) classification model capable of distinguishing high-functioning and low-functioning families based on family communication, adaptability, emotional security, and conflict management.

Methods and Materials: This cross-sectional predictive study was conducted among 796 families residing in Portugal. Participants were recruited using a stratified multistage sampling procedure from urban and suburban regions. Family functioning served as the target classification variable and was assessed using the Family Assessment Device. Predictor variables included family communication measured by the Family Communication Scale, family adaptability assessed through the Family Adaptability and Cohesion Evaluation Scales IV, emotional security measured by the Security in the Family System Scale, and conflict management evaluated using the Conflict Resolution Styles Inventory. Data preprocessing included standardization, missing-value imputation, and feature preparation. Families were classified into high-functioning and low-functioning groups according to established Family Assessment Device cutoff criteria. An SVM classifier with radial basis function and linear kernels was developed and optimized using five-fold cross-validation and grid search procedures. Model performance was evaluated using accuracy, precision, recall, specificity, F1-score, area under the receiver operating characteristic curve (AUC), confusion matrix analyses, permutation importance, and SHAP analyses.

Findings: All predictor variables demonstrated significant positive associations with family functioning. The optimized SVM model achieved a testing accuracy of 89.37%, precision of 0.88, recall of 0.89, specificity of 0.88, and an F1-score of 0.88. Receiver operating characteristic analysis indicated excellent discriminative performance with an AUC of 0.93. Feature importance analyses revealed that emotional security was the strongest predictor of family functioning, accounting for 32.7% of model contribution, followed by family communication (28.4%), family adaptability (21.9%), and conflict management (17.0%). SHAP analyses further confirmed that higher levels of all four predictors increased the probability of classification into the high-functioning family group.

Conclusion: The findings demonstrate that communication, adaptability, emotional security, and conflict management constitute powerful indicators of family functioning and can be effectively utilized within machine learning frameworks to classify families with high accuracy. Emotional security emerged as the most influential predictor, highlighting its central role in healthy family systems. The results support the application of SVM-based predictive models as valuable tools for family assessment, early identification of at-risk families, and the development of targeted preventive and intervention programs.