1 October 2021 in Neural Networks
21 December 2022 in International Journal of Machine Learning and Cybernetics
13 September 2024 in IEEE Transactions on Knowledge and Data Engineering
Concept drift is an important characteristic and inevitable difficult problem in streaming data mining. Ensemble learning is commonly used to deal with concept drift. However, most ensemble methods cannot balance the accuracy and diversity of base learners after drift occurs, and cannot adjust adaptively according to the drift type. To solve these problems, this paper proposes a targeted ensemble learning (Targeted EL) method to improve the accuracy and diversity of ensemble learning for streaming data with abrupt and gradual concept drift. Firstly, to improve the accuracy of the base learners, the method adopts different sample weighting strategies for different types of drift to realize bidirectional transfer of new and old distributed samples. Secondly, the difference matrix is constructed by the prediction results of the base learners on the current samples. According to the drift type, the submatrix with appropriate size and maximum difference sum is extracted adaptively to select appropriate, accuracy and diverse base learners for ensemble. The experimental results show that the proposed method can achieve good generalization performance when dealing with the streaming data with abrupt and gradual concept drift.
20 January 2020 in IEEE Access
Interval-valued data (IVD) is a kind of data where each feature is an interval. The midpoint and boundary are the two commonly used methods for representing IVD. However, their structure information (such as location, size) may be incomplete because only midpoint or endpoint is adopted which will lead to poor results of data processing. To better depict the structural information of IVD, a unified representation frame (URF) for IVD is proposed. It not only takes into account the size and location information, but the relationship between them as well. This frame can also represent the midpoint and boundary methods in a unified way. Besides, symmetrical uncertainty (SU) is adopted to measure the relationship between features and classes quantitatively, and irrelevant features will be eliminated based on SU. The proposed URF_ SU is applied in some traditional classifiers like LIBSVM, CART Tree and KNN. The experimental results on synthetic and real-world datasets demonstrate that the proposed approach is more effective than other representation methods of IVD in classification tasks.
23 November 2022 in Knowledge and Information Systems
2 February 2023 in Information Sciences