Multiple Classes to Obtain Stability the Accuracy Value in Various Population Dataset of Gift Loyalty Card for Fraud Detection using Artificial Neural Network (ANN)
The company produce gift loyalty card to make payment of customer become easy and to maintain the loyal customer. The fraud loss in the gift loyalty card is quite big so the company needs to develop an effective fraud detection system using machine learning with affordable cost. Some researchers conducted experiments for fraud detection such as comparing some supervised learning algorithms, modify the population size of dataset for learning and testing, implement data pre-processing, etc. Their objective is to find better accuracy value as the result. In our proposed research, multiple classes are used to produce a simple dataset for learning, testing, and validation so that the accuracy value is stable and high for various population size of dataset. It can be obtained in the population size of learning dataset from 30% until 90%.
Keywords - Gift Loyalty Card (GLC), Point of Sales (POS), Fraud Detection, Retail, Redeem or Swipe, Top-up, Activation, check Balance, Feature Extraction, Feature Selection, History Fraud Pattern, Artificial Neural Network (ANN), Genuine transaction, Fraud Indication, Back of House (BOH), First Data (FD), Multiple Classes, Dataset.