计算机科学与探索 ›› 2018, Vol. 12 ›› Issue (11): 1815-1826.DOI: 10.3778/j.issn.1673-9418.1710047

• 人工智能与模式识别 • 上一篇    下一篇

面向噪声数据的强化模糊规则模型及实现

贾海宁,王士同   

  1. 1. 江南大学 数字媒体学院,江苏 无锡 214122
    2. 江南大学 江苏省媒体设计与软件技术重点实验室,江苏 无锡 214122
  • 出版日期:2018-11-01 发布日期:2018-11-12

Reinforced Rule-Based Fuzzy Models for Noisy Data and Its Implementation

JIA Haining, WANG Shitong   

  1. 1. School of Digital Media, Jiangnan University, Wuxi, Jiangsu 214122, China
    2. Key Laboratory of Media Design and Software Technology of Jiangsu Province, Jiangnan University, Wuxi, Jiangsu 214122, China
  • Online:2018-11-01 Published:2018-11-12

摘要:

针对强化模糊规则模型抗噪性能较差的问题,提出了基于中智模糊聚类的强化模糊规则(reinforced rule-based fuzzy model based on neutrosophic C-means clustering,NCM-RRbF)模型。首先将中智模糊集与模糊C均值方法相结合,得到中智模糊聚类算法(neutrosophic C-means clustering,NCM)。然后通过将NCM方法用于强化模糊规则(reinforced rule-based fuzzy,RRbF)模型的初始规则生成,将基于背景的模糊聚类(context fuzzy C-means clustering,CFCM)方法用于RRbF模型的新规则生成,从而得到NCM-RRbF模型。该模型具有良好的降噪效果,适用于具有边界数据点和噪声数据点的聚类问题。通过对人工数据集及真实数据集添加不同噪声强度的高斯白噪声进行系统性实验,充分表明了所提出模型对于含有噪声和边界点的数据场景具有显著的逼近性能和良好的抗噪能力。

关键词: 模糊规则, 中智模糊聚类, 基于背景的模糊聚类, 加权最小二乘方法

Abstract:

Reinforced rule-based fuzzy model has obvious limitation for noisy data sets. To solve this drawback, a reinforced rule-based fuzzy model based on neutrosophic C-means clustering (NCM-RbF) is proposed. First of all, neutrosophic C-means clustering (NCM) algorithm is inspired from fuzzy C-means and the neutrosophic set framework. The NCM clustering algorithm is used to generate the initial rules of the reinforced rule-based fuzzy (RRbF) models, then the context fuzzy C-means clustering (CFCM) algorithm is used to generate the new rules of the RRbF models, and the reinforced rule-based fuzzy based on neutrosophic C-means clustering (NCM-RRbF) model is obtained. The proposed model has satisfactory anti-noise effect, which is suitable for clustering problem with boundary data points and noisy data points. For a systematic experiment of adding Gauss white noise with different noise intensity to artificial data sets and real data sets, the experimental results show that the proposed model has promising approximation capability and strong robustness for noisy data modeling tasks.

Key words: fuzzy rule, neutrosophic C-means clustering, context fuzzy C-means clustering, weighted least square estimation