Decision making using multiple models

Many real world business situations require classification decisions that must often be made on the basis of judgment and past performance. In this paper, we propose a decision framework that combines multiple models or techniques in a complementary fashion to provide input to managers who make such decisions on a routine basis. We illustrate the framework by specifically using five different classification techniques – neural networks, discriminant analysis, quadratic discriminant analysis (QDA), k-nearest neighbor (KNN), and multinomial logistic regression analysis (MNL). Application of the decision framework to an actual retail department store data shows that it is most useful in those cases where uncertainty is high and a priori classification cannot be made with a high degree of reliability. The proposed framework thus enhances the value of exception reporting, and provides managers additional insights into the phenomenon being studied.