Journal of Frontiers of Computer Science and Technology ›› 2021, Vol. 15 ›› Issue (1): 132-140.DOI: 10.3778/j.issn.1673-9418.1909021

• Artificial Intelligence • Previous Articles     Next Articles

Research on Dynamic Data Stream Classification Algorithm with New Class

WU Weijie, ZHANG Jingxiang   

  1. School of Science, Jiangnan University, Wuxi, Jiangsu 214122, China
  • Online:2021-01-01 Published:2021-01-07



  1. 江南大学 理学院,江苏 无锡 214122


Aiming at the low performance in detecting new class of classification algorithm on dynamic data stream with new class, a completely randomized forest algorithm based on k-nearest neighbor (KCRForest) is proposed. The algorithm constructs completely randomized trees of completely randomized forest by only known-class sam-ples in dynamic data stream, and divides the sample space into normal or abnormal region according to the average path length of leaf nodes. The outlier of a sample is obtained based on its k-nearest neighbor, when the sample falls into abnormal region. If the outlier is greater than the set threshold, the sample is judged to be new-class. Otherwise it is judged to be known-class. When the known-class sample falls into abnormal region, class distribution is obtai-ned based on its k-nearest neighbor. Otherwise class distribution can be obtained during training period. The label of known-class sample is identified by voting. When a certain number of new class samples are detected, the model is updated by the new-class sample information to detect other new classes. In order to verify the effectiveness of KCRForest algorithm in detecting new classes, experiments are carried out on 4 UCI datasets respectively, and comparisons are made with existing algorithms. The results show that the proposed method is equivalent to or better than iForest+SVM and LOF+SVM on new-class detection, and its classification accuracy is better than SENCForest.

Key words: new-class detection, completely randomized forest, dynamic data stream



关键词: 新类检测, 完全随机森林, 动态数据流