Efficient Resampling for Fraud Detection During Anonymised Credit Card Transactions with Unbalanced Datasets

Abstract

The rapid growth of e-commerce and online shopping have resulted in an unprecedented increase in the amount of money that is annually lost to credit card fraudsters. In an attempt to address credit card fraud, researchers are leveraging the application of various machine learning techniques for efficiently detecting and preventing fraudulent credit card transactions. One of the prevalent common issues around the analytics of credit card transactions is the highly unbalanced nature of the datasets, which is frequently associated with the binary classification problems. This paper intends to review, analyse and implement a selection of notable machine learning algorithms such as Logistic Regression, Random Forest, K-Nearest Neighbours and Stochastic Gradient Descent, with the motivation of empirically evaluating their efficiencies in handling unbalanced datasets whilst detecting credit card fraud transactions. A publicly available dataset comprising 284807 transactions of European cardholders is analysed and trained with the studied machine learning techniques to detect fraudulent transactions. Furthermore, this paper also evaluates the incorporation of two notable resampling methods, namely Random Under-sampling and Synthetic Majority Oversampling Techniques (SMOTE) in the aforementioned algorithms, in order to analyse their efficiency in handling unbalanced datasets. The proposed resampling methods significantly increased the detection ability, the most successful technique of combination of Random Forest with Random Under-sampling achieved the recall score of 100% in contrast to the recall score 77% of model without resampling technique. The key contribution of this paper is the postulation of efficient machine learning algorithms together with suitable resampling methods, suitable for credit card fraud detection with unbalanced dataset.

A Deep Learning Ensemble With Data Resampling for Credit Card Fraud Detection

“…This section compares the proposed method with wellperforming methods in recent scholarly articles, including a weighted extreme learning machine (Weighted ELM) [91], an optimized [92], a deep neural network (DNN) based classifier [93], a cost-sensitive SVM [94], a neural network ensemble [95], a random forest-based genetic algorithm wrapper method (GA-RF) [42], a method that sequentially combines the C4.5 and naïve Bayes classifiers [96], a dynamic weighted ensemble technique using Markov Chain [97], a model developed using random forest algorithm and SMOTE based resampling (RF-SMOTE) [98], an XGBoost model with SMOTE based resampling [99], an LSTM ensemble with SMOTE-ENN [9], a comparison of SMOTE and ADASYN based resampling with a DNN classifier [4], and an ANN model with random undersampling (RUS) The stacking-based DL ensemble obtained optimal performance in comparison with other well-performing methods in Table 3, reflecting the proposed method's robustness. Meanwhile, it would be beneficial to observe how the proposed approach would perform using a different dataset.…”

Machine Learning Approach in Predicting Fraudulent Job Advertisement

“…Several studies have explored imbalanced datasets in the context of different fraudulent cases, utilizing various resampling techniques and evaluation metrics (Rubaidi et al, 2022;Chen et al, 2021;Li et al, 2021;Mrozek et al, 2020;Bauder et al, 2018). Among the techniques used for handling imbalanced data were Random Undersampling (RUS), Random Oversampling (ROS), SMOTE, Borderline-SMOTE, Adaptive Synthetic Sampling (ADASYN), and cost-sensitive learning.…”

A Neural Network Ensemble With Feature Engineering for Improved Credit Card Fraud Detection

“…However, it is necessary to compare our approach with existing credit card fraud detection methods in the literature. The methods include the following: the sequential combination of C4.5 decision tree and naïve Bayes (NB) [5], a light gradient boosting machine (LightGBM) with a Bayesian-based hyperparameter optimization algorithm [14], a light gradient boosting machine (LightGBM) with a Bayesian-based hyperparameter optimization algorithm [14], a cost-sensitive SVM (CS SVM) [6], an optimized random forest (RF) classifier [34], a deep neural network (DNN) [35], a random forest classifier with SMOTE data resampling [36], an improved AdaBoost classifier with principal component analysis (PCA) and SMOTE method [37], a cost-sensitive neural network ensemble (CS-NNE) [38], a stochastic ensemble classifier operating in a discretized feature space [39], a model based on overfitting-cautious heterogeneous ensemble (OCHE) [40], a dynamic weighted ensemble technique using Markov Chain (DWE-MC) [41], and an extreme gradient boosting (XGBoost) ensemble classifier with SMOTE resampling technique [42].…”

“…Method Sensitivity Specificity AUC Kalid et al [5] C4.5+NB 0.872 1.000 -Taha et al [14] LightGBM --0.928 Makki et al [6] CS SVM 0.650 -0.620 Khatri et al [34] Optimized Random forest 0.782 --Alkhatib et al [35] DNN 0.955 -0.990 Mrozek et al [36] Random forest + SMOTE 0.829 -0.910 Zhou et al [37] AdaBoost + SMOTE + PCA --0.965 Yotsawat et al [38] CS-NNE -0.936 0.980 Carta et al [39] Stochastic Ensemble Classifier 0.915 -0.876 Xia et al [40] OCHE --0.937 Feng et al [41] DWE-MC --0.66 Xie et al [42] XGBoost…”