Predicting Divorce Prospect Using Ensemble Learning: Support Vector Machine, Linear Model, and Neural Network.

Area of Science:

• Social Sciences
• Computer Science
• Psychology

Background:

• Global divorce rates have significantly increased since 1970.
• Over half of marriages in the United States may end in divorce or separation.
• Predicting divorce is complex, involving numerous social and personal factors.

Purpose of the Study:

• To develop and evaluate a novel ensemble learning technique for divorce prediction.
• To identify key indicators and significant factors contributing to divorce.
• To assess the performance of machine learning algorithms in predicting marital dissolution.

Main Methods:

• An ensemble learning technique combining Support Vector Machine (SVM), passive aggressive classifier, and neural network (MLP).
• Creation of a question-based dataset by field specialists to gather marital information.
• Application of 5-fold cross-validation for robust performance evaluation.

Main Results:

• Achieved 100% accuracy in divorce prediction.
• Receiver Operating Characteristic (ROC) curve accuracy, recall, precision, and F1 scores were all approximately 97%.
• Identified critical indicators and significant factors influencing divorce prediction.

Conclusions:

• The proposed ensemble learning model demonstrates high efficacy in predicting divorce.
• Machine learning, particularly SVM, passive aggressive classifier, and MLP, can accurately identify individuals at risk of divorce.
• The study provides valuable insights into the factors most significant in marital dissolution.