2019 International Conference on Computing, Power and Communication Technologies (GUCON) · 2019 · 11 citations · 10 references
ComputingData ScienceBankingData MiningMachine LearningClassificationDecision TreesEnsemble LearningEnsemble MethodsFraud DetectionSupport Vector MachinesBoostingClass ImbalanceClass Imbalance HandlingClass Imbalance LearningCredit Card Fraud DetectionImbalanced ClassificationImbalanced LearningRandom Forest Classification
Fraud detection is an important application area of data mining. Due to the development in technology, there is an increase in the number of frauds nowadays. Detecting the frauds is one of the greatest challenges in organizations. A major challenge that comes while handling frauds in datasets is that the datasets are highly imbalance in nature. The fraud instances are below 1% as compared to normal transactions. In this paper, our focus is on the comparison of various available balancing techniques in conjunction with classifiers and identified the best combination. The dataset taken is the standard credit card data of a European bank. We have applied different balancing techniques like Down Sampling, Up Sampling, Regular SMOTE, Borderline SMOTE, SVM SMOTE and ADASYN. For comparison purpose, we have taken classifiers as bagging and boosting models. Results have shown that balancing dataset using SVM SMOTE followed by Random Forest classifier gave the best results with F-score value of 0.85.
10
SMOTE: Synthetic Minority Over-sampling Technique
Nitesh V. Chawla, Kevin W. Bowyer, Lawrence Hall et al. · Journal of Artificial Intelligence Research · 2002 · 31.2K citations · Full text
ADASYN: Adaptive synthetic sampling approach for imbalanced learning
Haibo He, Yang Bai, Edwardo A. Garcia et al. · 2008 · 4.4K citations
Sequence classification for credit-card fraud detection
Johannes Jurgovsky, Michael Granitzer, Konstantin Ziegler et al. · Expert Systems with Applications · 2018 · 483 citations
Computing, Bioinformatics, Finance +14
MSMOTE: Improving Classification Performance When Training Data is Imbalanced
Shengguo Hu, Yanfeng Liang, Lintao Ma et al. · 2009 · 263 citations