An Imbalanced Financial Fraud Data Model Based on Improved XGBoost and RUS Boost Fusion Algorithm with Pairwise
Junhao Xian · BCP Business & Management · 2023
As the socio-economic landscape evolves, the investigation into anti-fraud behaviors in shopping gains increasing significance. Although prior studies have utilized machine learning to tackle this issue, they often grapple with two key obstacles. First, an imbalance between positive and negative data samples exists. Second, the presence of redundant features leads to suboptimal model performance. In order to surmount these challenges, we've developed a new machine learning framework. This innovative solution automatically selects features and balances the data set's positive and negative samples. Our framework's outstanding performance on the IEEE-CIS Fraud Detection dataset thoroughly validates the efficacy of our approach.