FRAUD E-CHANNEL DETECTION WITH K-NEAREST NEIGHBORS (KNN) ALGORITHM

• Main Author: Purba, Adong
• Format: Theses
• Language: Indonesia
• Online Access: https://digilib.itb.ac.id/gdl/view/46585
• Institution: Institut Teknologi Bandung

Banking is the sector most often the victim of fraud using e-channel transactions, one of which is using an Automatic Teller Machine (ATM). Fraud is an act of deviation or negligence that is intentionally carried out to deceive or manipulate customers, or other parties, that occur in banks or use bank facilities so as to cause other parties to suffer losses and fraudsters get direct or indirect financial benefits. To control fraud, banks are required to have and implement an effective anti-fraud strategy by analyzing transaction data to look for suspicious patterns so as to facilitate identification of transactions as legitimate transactions or not. The classification method in the K-Nearest Neighbors (KNN) algorithm can be used to predict fraudulent transactions without the need to make a model first when detecting fraud. The results of KNN algorithm testing in the case of fraud detection get an ROC AUC score of 0.71 and an accuracy of 66%. Because the ROC AUC value is in the range of 0.7-1.0, the method is included in the quite good category based on data mining classification.