DIVORCE PREDICTION USING CORRELATION BASED FEATURE SELECTION AND ARTIFICIAL NEURAL NETWORKS

摘要: Within the scope of this research, divorce prediction was carried out by using Divorce Predictors Scale (DPS) on basis Gottman couples therapy. Of participants, 84 (49%) were divorced and 86 (51%) married couples. Participants completed “Personal Information Form” “Divorce Scale”. In study, success DPS, investigated Multilayer Perceptron Neural Network C4.5 Decision tree algorithms. addition, study also aims to find most significant features/items in that affect divorce. The effective 6 features their values significance obtained applying correlation-based feature selection method data set. When we look at these features, they are related creating a common meaning failed attempts repair, love map negative conflict behaviors. direct classification methods applied set, highest rate 98.23% with RBF neural network. After selecting same accuracy 98.82% ANN. According results, DPS can predict Family counselors family therapists use scale for contribute preparation case formulation intervention plan. Also it be said predictors therapy confirmed Turkish sampling.