Modelo matemático para predecir los divorcios en la ciudad de Latacunga

Abstract

Currently, the early prediction of a divorce case is essential to save the marriage. The sample selected for this study was 381 people between married and divorced couples from urban parishes in the city of Latacunga. The data was obtained through the survey, through the questionnaire instrument that uses Professor Gottman's method. To determine the most important variables or characteristics that can lead to a divorce, the free software Rstudio was used. In a first sweep using the variable correlation method, 15 variables were obtained, then in a second sweep using the WEKA software, 5 variables or characteristics were obtained. The independent variables are: Atr2(Try to ignore differences even in difficult situations), Atr3(Restart arguments from the beginning and correct them), Atr10 (Most of our goals are unusual) Atr35(I insult my spouse in an argument), Atr43 (I remain silent to calm the atmosphere in an argument), and divorce was the dependent variable. Two supervised classification algorithms were used: Logistic Regression and Naive Bayes, executing the predictions with 70% training data and 30% testing data. To evaluate the performance in the classification, the precision and F1_Score metrics were used. The tests determined that the value of the F1-Score of the Logistic Regression model is (98%) with an accuracy of (98%), while the F1Score value of the Naive Bayes model is (96%) and its accuracy of (96%). Concluding that the Logistic Regression model is the best model to predict divorces in the city of Latacunga.