Implementation of Classification Algorithm for Anomaly Detection in Credit Card Transactions with RapidMiner

This study evaluates the application of the Random Forest algorithm in the classification of credit card transactions to detect transaction types such as cash_out, cash_in, payment, transfer, and debit. The dataset used is derived from Kaggle.com and includes attributes such as the number of transactions, sender and recipient balances, and transaction types. The results showed an accuracy of 80.63%, with the best performance in cash_out and payments, but difficulties in classifying debit and transfer transactions due to class imbalances. Classroom balancing using smote or undersampling, as well as unsupervised learning techniques, can improve model performance. Improving the model through feature engineering and hyperparameter tuning is also needed to improve the effectiveness of fraud detection.