چكيده به لاتين
Annually, billions of dollars are lost due to the fraudulent transactions made with credit cards. The design of an algorithm and an effective method is the key to solving this crisis and financial losses. Because of speeding up the works and the ease of use, Credit Cards are very common and widely used. The credit card is used both by legal users and by fraudsters. Fraudsters use new and different ways to infiltrate and threaten E-commerce systems, these threats from these groups have led to the creation of fraud detection systems. Fraud Detection algorithms by analyzing user transaction data can disclose fraudulent behaviors and activities in various E-commerce areas. In this research, we are developing an algorithm based on data mining methods seeking to detect fraud in credit card transactions.
Fraud detection algorithms have problems and defects, which makes it impossible to make the right decision about whether a transaction is normal or fraudulent. These problems include unbalanced databases, unrelated features, the high number of false alarms, and the use of a similar data analysis model in the detection method. The algorithm presented in this study is based on the ensemble learning. This will increase the accuracy and precision of the results.
Using ensemble learning algorithm to detect fraud in credit card transactions compared with previous algorithms resulted in 100% precision, 99% accuracy, and 97.5% recall. The research was conducted on two-day transactions of a European bank.
