EcoSta 2022: Start Registration
View Submission - EcoSta2022
A0474
Title: A novel approach to bank marketing campaign Authors:  Yuzhi Cai - Swansea University (United Kingdom) [presenting]
Abstract: Banks usually need to identify a group of customers in order to target them with a specific financial product that will allow them to retain existing and attract new customers. A binary response model could be used to predict the response probability for each customer, which could then be used for classification. However, as the predicted response probability depends on the mean of an unobserved variable, it is not clear whether these response probability forecasts are able to provide optimal classification results for, e.g. bank marketing campaigns. We show that these response probability forecasts usually do not lead to optimal classification results because they use only one specific piece of information about the response variable and this piece of information may not be representative of the response variable. We also propose a novel binary quantile function model and a new classification method that allow us to use a set of information about the response variable from which optimal classification results can be obtained. We illustrate this by analysing some real bank marketing data.