Due to regurations and laws prohibiting uses of private data on customers and their transactions in customer data base, most customer data sets are not easily accessable even in the same organizations. A solutio for this reguatory problems can be providing statistical summary of the data or models induced from the dat, instead of providing raw data sets. The models, however, have limited information on the original raw data set. This study explores possible solutions for these problems. The study uses prediction models from data on credit information of customers provided by a local bank in Seoul, S. Korea. This study suggests approaches in figuring what is inside of the non-rules based models such as regression models or neural network models. The study proposes several rule accumulation algorithms such as (RAA) and a GA-based rule refinement algorithm (GA-RRA) as possible solutions for the problems. The experiments show the performance of the random dataset, RAA, elimination of redundant rules (ERR), and GA-RRA.
Personal credit ratings, logistic regression model, rule accumulation algorithm (RAA), GA-based rule refinement algorithm (GA-RRA)
ISBN: [978-1-86435-644-1]; Full paper
Kim, Jinhwa; Bae, Jae Kwon; and Delen, Dursun, "Understading Black Boxes: Knowledge Induction From Models" (2011). PACIS 2011 Proceedings. 97.