Incremental Learning Algorithm of Data Complexity Based on KNN Classifier

作者：Li Jie, Yaxu Xue, Yadong Yu · 年份：2020 · DOI：10.1109/ccs49175.2020.9231514 · 被引用次数：3 · 研究领域：Machine Learning and Data Classification、Imbalanced Data Classification Techniques、Data Stream Mining Techniques

In practice, new data are constantly generated, and the existing data complexity algorithms are based on the idea of batch learning. In the face of the dynamic increase of data scale, how to measure the characteristic information of data has become an urgent problem to be solved in the field of data mining. This paper focuses on this problem and further studies its incremental learning function on the basis of in-depth discussion of data complexity proposed by TK Ho et al. Among them, N3 and N4 are the complexity indexes based on KNN classifier (K=1). In this paper, the incremental learning algorithm I1NN was proposed on the basis of 1-NN classifier, and its feasibility and validity were verified on both the artificial data set and the UCI public data set.