<![CDATA[Anomalies in bank transaction data often indicate fraudulent activity or errors. This research aims to detect anomalies in bank transaction data by optimizing DBSCAN parameters using the Differential Evolution (DE) method because there are shortcomings, namely the difficulty of determining the right parameters to create the right cluster in order to detect anomalies in bank transaction data properly. The data used is transaction data from Bank XYZ with more than 1011 data records. The research stages include data collection, data preprocessing (data cleaning, normalization, and transformation), system design, algorithm implementation, and analysis and testing using the Silhouette score and Z-score methods. The DE method is used to automatically determine the optimal parameters of MinPts and Epsilon. The results show that the use of DE can produce optimal parameters, with increased anomaly detection accuracy using DBSCAN. Evaluation with Silhouette score shows an average accuracy of 0.7916 and using DBI reaches 0.19791 at the lowest, while Z-score and MSE measurements show high cluster density with anomaly detection accuracy reaching 98.41% and 0.555537. The DE approach to parameter selection is effective in improving the performance of DBSCAN in detecting anomalies in bank transaction data. Suggestions for future research are to increase the number of data records and conduct experiments on a wider variety of data attributes.]]>
Copyrights © 2024
