Marital Stability and Divorce Prediction Among Couples: A Machine Learning Approach

Marital dissolution usually affects one’s health destructively. This paper aims to examine significant factors in predicting marital stability among Iranian couples with high accuracy using machine learning algorithms. Machine learning algorithms can investigate significant features in marital stability, identify the most important features in such complicated information with multivariate data, and predict marital stability with high accuracy. Firstly, Young Schema Questionnaire–Short Form 2, Persian Divorce Prediction Scale, and Gottman Emotional Separation Scale are completed by participants. Then, two feature selection techniques, namely Univariate and Relief, are used. Afterward, four machine learning practices, including support vector machine (SVM), logistic regression (LR), decision tree (DT), and random forest (RF), are trained and tested on the set of full features alongside the selected features for their ability to predict marital stability with the highest possible accuracy. Several metrics are employed to evaluate the performance of multi-class classification predictors. After hyperparameter tuning, SVM and RF training algorithms achieved more than ‏89% accuracy on the created dataset. The findings can be utilized both by couples therapists and couples vulnerable to marital dissolution.