Credit Card Fraud Identification Based on Principal Component Analysis and Improved Adaboost Algorithm
Hua Zhou, Luyang Wei, Gangyi Chen, Peng Chun Lin, Yangkai Lin · 2019
As the issuance of credit cards grows rapidly, its risk cannot be underestimated. Due to the high dimension and sparsity nature of credit cards fraud data, it is easy to occur overfitting, low robustness and poor generalization in data mining. Therefore, this paper proposes a risk control algorithm for integrated learning, which is based on combining smote and principal component analysis for data processing, by using a single-layer decision tree and the improved Adaboost algorithm, as well as applying the trained risk control system to individual data for fraud identification. After testing the data sample of a commercial bank, the simulation results show that the accuracy rate of the improved integrated machine learning method reaches 96.50% and F-Measure value is 97.3%. In addition, comparing with other detection methods, the result proved that the proposed method is superior to traditional risk control technology, which also effectively completed the fraud detection work.