Diagnosis of internal frauds using extreme gradient boosting model optimized with genetic algorithm in retailing

Abstract

Fraud is one of the most vital problems that can lead to a loss of organizational reputation, assets and culture. It is beneficial for companies to anticipate possible fraud in order to protect both culture and company assets. The aim of this study is to provide a fraud detection model using classification and optimization algorithms. For this purpose, this study proposes a novel hybrid model called XGBoost-GA to enhance the prediction quality for cashier fraud detection in retailing. In the proposed model, the genetic algorithm (GA) is used to optimize the parameters of extreme gradient boosting (XGBoost) model. The proposed XGBoost-GA model is compared with XGBoost, logistic regression (LR), naive bayes (NB) and k-nearest neighbor (kNN) algorithms. The performance comparison is presented with a case study with the actual data taken from a grocery retailer in Turkey. Numerical results showed that the proposed hybrid XGBoost-GA model produces higher accuracy, recall, precision and F-measure than other classification algorithms. In this context, the use of proposed model in fraud detection will be beneficial for companies to use their resources effectively. Classification algorithms will also accelerate organizations in terms of detecting the possible damage of fraud to company assets before it grows.