Prediction Divorce Prospect using Machine Learning Methods: A Case Study on Bangladesh

Abstract

In this paper, we suggest a system for predicting divorce and use the Divorce Predictors Scale to evaluate it (DPS). The DPS is a 40-item self-report questionnaire based on Gottman couple's therapy, which could be used as features or characteristics in a machine learning algorithm. Besides from "Divorce Predictors Scale," a "personal information form" was utilized to collect personal information from participants in a more traditional and disciplined manner. There were 179 (71.6%) married people and 71 (28.4%) divorced people among the 250 participants (N=250). The algorithms of K-nearest neighbors, Naive Bayes, and Decision Tree, SVM, Random Forest were used. We attempted to restrict the field using feature selection. a list of the most important or noteworthy characteristics based on the selection of features based on correlation as a result, characteristics/items were discovered. In the context of Bangladeshi divorce prediction, data. Different algorithms are applied directly to the data throughout this process. The dataset with the greatest accuracy rate for divorce prediction is the SVM method produced an accuracy of 100%. The two algorithm Decision Tree, Naïve Bayes given the same accuracy 94.0%. However, the Random Forest which give accuracy 96.0%. DPS can predict divorce, as evidenced by the results. This scale can be used by family advocates and family experts to help with case characterization and mediation strategy. Furthermore, the results of the study show that the Machine Learning Algorithm treatment verified in the Bangladeshi sample.