The usage of statistical analysis to predict credit card acceptance and income level

This paper presents a statistical analysis of variables within a credit card approval dataset to predict approval decisions and income level of applicants. Descriptive analysis was performed on continuous variables and inferential statistical techniques, including confidence intervals, t-tests, chi-square tests, and ANOVA, were employed to predict credit card approval. Using a binary logistic regression model, a misclassification rate of 0.48% was achieved. As for predicting income level, Various models were developed and evaluated to identify the most effective approach. Findings indicate that Model 5, which incorporated data centering and adjusted reference levels for categorical variables, demonstrated superior performance by effectively mitigating multicollinearity, though it still had a low R² of 20.8%, MAPE of 32.3% and RAE or 0.86.