论文部分内容阅读
针对传统分类器的泛化性能差、可解释性及学习效率低等问题,提出0阶TSK-FC模糊分类器.为了将该分类器应用到大规模数据的分类中,提出增量式0阶TSK-IFC模糊分类器,采用增量式模糊聚类算法(IFCM(c+p))训练模糊规则参数并通过适当的矩阵变换提升参数学习效率.仿真实验表明,与FCPM-IRLS模糊分类器、径向基函数神经网络相比,所提出的模糊分类器在不同规模数据集中均能保持很好的性能,且TSK-IFC模糊分类器在大规模数据分类中尤为突出.
To overcome the problems of poor generalization performance, interpretability and low learning efficiency of traditional classifiers, a 0-order TSK-FC fuzzy classifier is proposed.In order to apply this classifier to the classification of large-scale data, an incremental 0-order IFCM (c + p)) is used to train the parameters of fuzzy rules and to improve the efficiency of parameter learning through appropriate matrix transformation. Simulation results show that, compared with FCPM-IRLS fuzzy classifier, Compared with radial basis function neural network, the proposed fuzzy classifier can maintain good performance in different scale datasets, and the TSK-IFC fuzzy classifier is particularly prominent in large-scale data classification.