A new over-sampling ensemble approach for imbalanced data
Peng Liu, Xiaoyuan Liu, Bo Liu, Xinyi Chen · 2021
With the rapid development of information technology, data in various fields is being generated, collected and stored at an unprecedented speed. With the increase in the amount of information, problems such as data imbalance and noise have become increasingly serious. However, traditional classification algorithms aiming at maximizing the overall classification accuracy tend to favor the majority class while ignoring the minority class when dealing with imbalanced data. In order to solve this problem, this paper proposes a new algorithm, which combines oversampling technology with boosting ensemble learning algorithm, and uses the error function of each round to guide the next round of sampling procedures. Experiments have proved that our algorithm has higher classification accuracy than other widely-used imbalanced learning methods, and has better adaptability because it does not depend on the specific classifier.