The discriminant analysis technique works well only when the set of independent variables/covariates follows multivariate normal distribution. At the starting point, it does not show flexibility by ruling out categorical variables. Many economic variables, such as salary, savings, and so on, are known not to follow normal distribution and are also skewed in general. Thus, the assumption of multivariate normal distribution is rather restrictive and we need a general framework for the classification problem. A very important class of model is provided by the logistic regression model. In fact, it is known to have very nice theoretical properties. For example, it is theoretically known that the logistic regression model provides as much accuracy as the discriminant analysis in the case of the independent variable following the multivariate normal distribution. The logistic regression model is a member of the important exponential family, and it belongs...