Student: Cerlang Gemintang (IUP – 15/386838/PA/17049), Supervised by : Dr. Mardhani Riasetiawan. International Undergraduate Program, Computer Sciences, Department of Computer Science and Electronic, Faculty of Mathematics and Natural Sciences, Universitas Gadjah Mada. Presented at July 1, 2019.
Credit card fraud can be defined as deceptive or unauthorised use of another person’s credit card account in an attempt to steal money, goods, or services. Credit card fraud is a major problem, especially in developing countries and causes millions of dollars in losses annually. A credit card fraud detection system would help greatly in combatting this threat and reducing the damage it causes.
This research focuses on the use of DBSCAN is order to perform outlier detection on a dataset for the purposes of detecting fraud. The system was experimented on with various parameters in addition with data transformations, and then compared with another well-known clustering algorithm, K-means in order to determine the effectiveness of its performance.
The results of the research showed that DBSCAN was able to more accurately determine fraudulent transactions than K-means. In addition, the research showed that utilising PCA transformation on the dataset first improves accuracy of the clustering and thus of outlier detection.