Predicting Marital Satisfaction Using Random Forest Based on Attachment, Communication, Intimacy, and Conflict Indicators

Abstract

Objective: This study aimed to predict marital satisfaction among married adults in Canada using Random Forest regression based on attachment, communication, intimacy, and conflict indicators.

Methods and Materials: This cross-sectional predictive study was conducted on 646 married adults residing in Canada. Participants completed the Couples Satisfaction Index-16, Experiences in Close Relationships–Revised questionnaire, Communication Patterns Questionnaire–Short Form, Personal Assessment of Intimacy in Relationships scale, and selected dimensions of the Revised Conflict Tactics Scales. The predictor set included attachment anxiety and avoidance, constructive communication, mutual avoidance, demand-withdraw communication, emotional, social, sexual, intellectual, and recreational intimacy, negotiation, and psychological aggression. Random Forest regression was used to predict continuous marital satisfaction scores. The dataset was divided into training and independent test subsets using an 80:20 ratio, and hyperparameters were optimized through five-fold cross-validation. Predictive performance was evaluated using R², mean absolute error, mean squared error, and root mean squared error. Permutation importance and partial dependence analyses were used to interpret predictor contributions and nonlinear patterns.

Findings: The Random Forest model demonstrated strong predictive performance, with an R² of .91 in the training set, .76 in five-fold cross-validation, and .78 in the independent test set. Test-set prediction error was MAE = 5.06, MSE = 43.52, and RMSE = 6.60. Emotional intimacy showed the greatest permutation importance, followed by constructive communication, attachment anxiety, sexual intimacy, psychological aggression, attachment avoidance, negotiation, and demand-withdraw communication. Partial dependence analyses indicated that higher emotional intimacy, constructive communication, sexual intimacy, and negotiation were associated with higher predicted marital satisfaction, whereas attachment anxiety, psychological aggression, attachment avoidance, and demand-withdraw communication were associated with lower predicted satisfaction.

Conclusion: Marital satisfaction can be predicted with substantial accuracy using a multidimensional Random Forest model, with emotional intimacy and constructive communication emerging as the strongest positive predictors and attachment insecurity and destructive conflict as the principal negative predictors.