With the aim of creating risk-free online payment systems by improving banking system reliability, this study aims at creating credit card fraud detection system using automation approach. By reviewing existing literature, the study identifies inadequacies in current methodologies, particularly in dataset balancing, feature importance analysis, and model interpretability. This study investigates advanced supervised machine learning techniques for credit card fraud detection using a dataset of 1,852,394 transactions. There are three algorithms—decision tree, logistic regression, and random forest—with a focus on RF due to its high accuracy and ability to prevent overfitting. By introducing innovative features such as category probability, vowel-consonant ratios, age probability, and geospatial data, to enhance the models’ capacity to identify fraudulent transactions. The models are examined using multiple metrics, including recall, precision, F1-score, and precision-recall area under curve (PR-AUC). The RF model demonstrates superior performance, highlighting the effectiveness of feature engineering and appropriate evaluation metrics in developing robust fraud detection systems. This research provides valuable insights for improving fraud prevention strategies.