Divorce Predictors Scale (DPS) on the basis of Gottman couples therapy was used to accurately model divorce prediction using a pre-existing data-set. There were 170 participants in the data-set which of them 84 were divorced and 86 married and they all completed “Personal Information Form” and “Divorce Predictors Scale”. We applied various machine learning approaches to build a predictive model which estimates the most accurate divorce prediction results. Feature selection process was used to find the most significant attributes by eliminating unnecessary and noise attributes which had negative impact during the training process. Classification methods were applied directly on the data-set and after the feature selection. After applying all the algorithms our best results were obtained by Linear Discriminant Analysis, K-Nearest Neighbours, Random Forest, and Support Vector Machine. Adding the feature selection methods to eliminate the unnecessary attributes and applying the algorithms, the accuracy and the Kappa value of the model increased for Logistic Regression, and Quadratic Discriminant Analysis, Random Forest.
The dataset used is available here at - https://archive.ics.uci.edu/ml/datasets/Divorce+Predictors+data+set