A data record of 70,000 individuals are taken to form training and predicting set to develop model for churning probability determination. It included use of sampling, factor analysis, cross-correlation and finally logistic regression.