Artificial Intelligence Framework for Divorce Risk Prediction and Early Marital Intervention.

Abstract

Divorce emerged as one of the significant social problems, posing psychological, economic, and family-related issues, resulting in complications. Therefore, there is a need to formulate reliable and valid techniques that enable early prediction and facilitate divorce prevention practices. This study introduces an intelligent and reliable cascaded hybrid Artificial Deep Neural Network–Random Forest (ADNN–RF) system which can effectively predict divorce risk based on socio-demographic data, conflict resolution tendencies and marital satisfaction to facilitate timely intervention and evidence-based marriage counseling. We employed quantitative survey research design and collected data from 402 married and divorced individuals using a structured Google Forms questionnaire. Twenty variables measuring respondents’ relationship attributes on a five-point Likert scale were used to train and test the proposed hybrid model. Further, descriptive statistics were used to understand the general tendencies of participants concerning constructive conflict resolution behaviours and relationship satisfaction. Pearson Chi-square test revealed that there exists a significant relationship between conflict resolution behaviours and divorce status (χ² = 12.84, p = 0.001), verifying our hypothesis that partners’ abilities to manage conflict predict divorce status. Results demonstrated that the proposed deep learning framework could be used by marriage counsellors, psychologists, divorce experts, and authorities as a decision- support system in building intelligence that predicts divorce and reduces its occurrence by triggering early warning signs and recommending timely interventions. The formulated hybrid model provides scientific-based evidence that can assist family counsellors in supporting couples to enhance healthy relationships and avoid divorce.