Self-service checkout is now increasingly used in supermarkets and retail. Customers scan all items with a barcode reader to buy and pay for them. This type of payment carries the risk of fraud by not scanning all items in the cart.. Food retailers hope to be able to identify cases of fraud to minimize losses. Data mining techniques and algorithms are used to check the classification of large amounts of information and data recovery. One group of these methods is the Anomaly Detection Method. The history of customer behavior is considered routine. If the new action is seen from customers who are not compatible with the movement pattern, the likelihood of fraud will be even more substantial. The aim is to predict fraud in one retail self check out using classification techniques and visualize the results to obtain new insight. The results of this study are oversampling on unbalanced data that can improve the performance of a model. The model with the best performance is J48 with F-measure 0.981. New insight from the visualization of prediction results is that customers with low trust levels are vulnerable to fraud.
Paper
Full text
Fraud Detection at Self Checkout Retail using Data Mining
Semantic Scholar · Computer Science · 2020
Abstract
Self-service checkout is now increasingly used in supermarkets and retail. Customers scan all items with a barcode reader to buy and pay for them. This type of payment carries the risk of fraud by not scanning all items in the cart.. Food retailers hope to be able to identify cases of fraud to minimize losses. Data mining techniques and algorithms are used to check the classification of large amounts of information and data recovery. One group of these methods is the Anomaly Detection Method. The history of customer behavior is considered routine. If the new action is seen from customers who are not compatible with the movement pattern, the likelihood of fraud will be even more substantial. The aim is to predict fraud in one retail self check out using classification techniques and visualize the results to obtain new insight. The results of this study are oversampling on unbalanced data that can improve the performance of a model. The model with the best performance is J48 with F-measure 0.981. New insight from the visualization of prediction results is that customers with low trust levels are vulnerable to fraud.