Analysis of Chi-square Automatic Interaction Detection (CHAID) for Segmentation and Customer Targeting to Minimize Non-performing Loan
Abstract
Banking plays an important role in the economy because it can increase
growth and development, especially in the economic sector. Bank is one of the
financial institutions or companies engaged in finance that carry out various kinds
of services, such as providing loans, distributing currencies, controlling currencies,
as a place to store valuable objects, finance companies, etc (Limbong et al. 2010).
Credit is the most important thing in bank activities, considering that the largest
income of a bank is obtained from the credit sector. CHAID analysis is conducted
to identify potential customer segments based on the variables that have the most
significant influence to credit status. SMOTE is conducted for managing
unbalanced data with 600% oversampling and 117% undersampling. Based on
CHAID analysis, the characteristics that characterize the variety of credit status
were the loan term, the latest education, gender, age, and loan amount with 5%
significant level. CHAID analysis produced 14 segments based on the
independent variable associated with the response variable which in this case the
response variable is credit status with a classification accuracy of 67.9%.