Credit card fraud has been a growing issue both in the United States and worldwide. Unfortunately, the countless losses caused by fraudulent transactions are usually paid for by the sales companies and credit card companies. This project analyzes supervised machine learning models for identifying fraudulent transactions based on the transaction information. In doing so, we have worked with three models: decision tree, KNN, and random forest, analyzing both the individual models and all possible combinations of models. A final fraud score is calculated as a probability of fraud based on all models, which achieved a recall of 82% with a precision of 84%. Our final model is a probability and not an absolute decision of fraud, which can allow credit card companies to factor in personal concerns when deciding their course of action in response to possible future frauds.
View Our Decision Tree Predictions