基于知识量加权的直觉模糊均值聚类方法

针对聚类算法中特征数据对聚类中心贡献的差异性及算法对初始聚类中心的敏感性等问题,提出一种基于知识量加权的直觉模糊均值聚类方法。首先将原始数据集直觉模糊化并改进最新的直觉模糊知识测度计算知识量,据此实现数据集特征加权,再利用核空间密度与核距离初始化聚类中心,以提高高维特征数据集的计算精度与聚类效率,最后基于类间样本距离与最小知识量原理建立聚类优化模型,得到最优迭代算法。基于UCI人工数据集的实验结果表明,所提方法较大程度地提高了聚类的准确性与迭代效率,分类正确率及执行效率分别平均提高了10.63%和31.75%,且具有良好的普适性和稳定性。该方法首次将知识测度新理论引入模糊聚类并取得优良效果,为该理论在其他相关领域的潜在应用开创了新例。     

基于特征加权距离的双指数模糊子空间聚类算法

传统的模糊聚类算法(FCM)使用欧氏距离计算数据点之间的差异时,对于高维数据集聚类效果不够理想.对此,以FCM算法的目标函数为基础,用特征加权距离代替传统的欧氏距离,同时向约束条件中引入指数γ和β,提出了一种基于特征加权距离的双指数模糊子空间聚类算法,并讨论了该算法的收敛性.实验表明,所提出算法可以有效提取高维数据集各类别的相关特征,在真实数据集上有较好的聚类效果.     

基于WFKP聚类算法的培养质量评估

针对传统培养质量评估方法过于依赖于专家系统,评估效率低、准确性不高的问题,提出一种特征加权的聚类算法(weighted fuzzy k-prototypes,WFKP)。在马氏距离中引入比例系数定义一种新的相异度,结合K近邻算法和簇内簇间离散度来计算数值属性的权值,利用互信息分析分类属性对聚类的依赖程度,提高聚类效率和准确性。通过UCI数据集验证了WFKP聚类算法的有效性和正确性,并应用到研究生培养质量评估中。实验结果表明,该算法可以挖掘提升培养质量的关键要素,避免评估过程的主观性。     

一种加权模糊C均值聚类算法及其在图像分割中的应用

现有的加权模糊C均值聚类算法中,属性加权是一个不断迭代、重复计算的过程,费时费力。针对这种情况,提出Fisher线性判别率进行属性加权。算法首先直接计算每一维属性对模糊聚类的贡献度,其次对所有属性的贡献度进行归一化处理然后加权聚类。在人工和实际数据集所做实验表明:该算法在提高聚类速度的同时,聚类效果上也优于其他同类加权模糊C均值聚类算法。     

基于样本-特征加权的可能性模糊核聚类算法

经典的模糊C-均值聚类算法存在对噪声数据较为敏感、未考虑样本属性特征间的不平衡性及对高维数据聚类不理想等问题,而可能性聚类算法虽然解决了噪声敏感和一致性聚类问题,但算法假定每个样本对聚类的贡献程度一样。针对以上问题,提出了一种基于样本-特征加权的可能性模糊核聚类算法,将可能性聚类应用到模糊聚类中以提高其对噪声或例外点的抗干扰能力;同时,根据不同类的具体特性动态计算样本各个属性特征对不同类别的重要性权值及各个样本对聚类的重要性权值,并优化选取核参数,不断修正核函数把原始空间中非线性可分的数据集映射到高维空间中的可分数据集。实验结果表明,基于样本-特征加权模糊聚类算法能够减少噪声数据和例外点的影响,比传统的聚类算法具有更好的聚类准确率。     

基于样本-特征加权的可能性模糊核聚类算法

经典的模糊C-均值聚类算法存在对噪声数据较为敏感、未考虑样本属性特征间的不平衡性及对高维数据聚类不理想等问题,而可能性聚类算法虽然解决了噪声敏感和一致性聚类问题,但算法假定每个样本对聚类的贡献程度一样。针对以上问题,提出了一种基于样本-特征加权的可能性模糊核聚类算法,将可能性聚类应用到模糊聚类中以提高其对噪声或例外点的抗干扰能力;同时,根据不同类的具体特性动态计算样本各个属性特征对不同类别的重要性权值及各个样本对聚类的重要性权值,并优化选取核参数,不断修正核函数把原始空间中非线性可分的数据集映射到高维空间中的可分数据集。实验结果表明,基于样本-特征加权模糊聚类算法能够减少噪声数据和例外点的影响,比传统的聚类算法具有更好的聚类准确率。     

不平衡数据加权集成学习算法

针对传统的机器学习算法对不平衡数据集的少类分类准确率不高的问题,基于支持向量机和模糊聚类,提出一种不平衡数据加权集成学习算法。首先提出加权支持向量机模型(Weighted Support Vector Machine,WSVM),该模型根据不同类别数据所占比例的不同,为各类别分配不同的权重,然后将WSVM与模糊聚类结合提出一种新的集成学习算法。将本文提出的算法应用于人造数据集和UCI数据集实验中,实验结果表明,所提出的算法能够有效地解决不平衡数据的分类问题,具有更好的分类性能。     

一种有效的分类型数据聚类方法

鉴于传统的K-means聚类算法只限于处理数值型数据,将K-means算法扩展到分类型数据域,提出一种分类型数据聚类方法.根据与每个分类属性的每个值相关的数据分布信息,同时结合数据的纵向与横向分布来评价数据对象与类之间的差异性,定义了一种新的距离度量.该方法能发现同一属性不同值间的内在关系,并能有效地度量对象间的差异性.用UCI中的数据集对所提算法进行验证,实验结果表明了该算法具有较好的聚类效果.     

基于自适应马氏距离的模糊c均值算法

经典的模糊c均值（FCM）算法是基于欧氏距离的,它只适用于球型结构的聚类,且在处理高维的数据集时,分错率增加。针对以上两个问题,提出了一种新的聚类算法（FCM-M）,它将马氏距离与模糊c均值相结合,并在目标函数中引进一个协方差矩阵的调节因子,利用马氏距离的优点,有效地解决了FCM算法中的缺陷,并利用特征值、特征矢量及伪逆运算来解决马氏距离中遇到的奇异问题。通过数据聚类和图像分割两组实验,证实了该方法的可行性和有效性。     

一种基于Mahalanobis距离的增量聚类算法

经典的模糊c均值聚类算法对非球型或椭球型分布的数据集进行聚类效果较差。将经典的模糊c均值聚类中的欧氏距离用Mahalanobis距离替代,利用Mahalanobis距离的优点,将其用于增量学习中,提出一种基于马氏距离的模糊增量聚类学习算法。实验结果表明该算法能较有效地解决模糊聚类方法中的缺陷,提高了训练精度。     

模糊C-均值聚类算法的优化

针对传统模糊C-均值聚类算法（FCM算法）初始聚类中心选择的随机性和距离向量公式应用的局限性,提出一种基于密度和马氏距离优化的模糊C-均值聚类算法（Fuzzy C-Means Based on Mahalanobis and Density,FCMBMD算法）。该算法通过计算样本点的密度来确定初始聚类中心,避免了初始聚类中心随机选取而产生的聚类结果的不稳定;采用马氏距离计算样本集的相似度,以满足不同度量单位数据的要求。实验结果表明,FCMBMD算法在聚类中心、收敛速度、迭代次数以及准确率等方面具有良好的效果。     

Mahalanobis distance based on fuzzy clustering algorithm for image segmentation

Conventional Fuzzy C-means (FCM) algorithm uses Euclidean distance to describe the dissimilarity between data and cluster prototypes. Since the Euclidean distance based dissimilarity measure only characterizes the mean information of a cluster, it is sensitive to noise and cluster divergence. In this paper, we propose a novel fuzzy clustering algorithm for image segmentation, in which the Mahalanobis distance is utilized to define the dissimilarity measure. We add a new regularization term to the objective function of the proposed algorithm, reflecting the covariance of the cluster. We experimentally demonstrate the effectiveness of the proposed algorithm on a generated 2D dataset and a subset of Berkeley benchmark images.     

A powerful hybrid clustering method based on modified stem cells and Fuzzy C-means algorithms

One of the simple techniques for Data Clustering is based on Fuzzy C-means (FCM) clustering which describes the belongingness of each data to a cluster by a fuzzy membership function instead of a crisp value. However, the results of fuzzy clustering depend highly on the initial state selection and there is also a high risk for getting the best results when the datasets are large. In this paper, we present a hybrid algorithm based on FCM and modified stem cells algorithms, we called it SC-FCM algorithm, for optimum clustering of a dataset into K clusters. The experimental results obtained by using the new algorithm on different well-known datasets compared with those obtained by K-means algorithm, FCM, Genetic Algorithm (GA), Particle Swarm Optimization (PSO), Ant Colony Optimization (ACO), Artificial Bee Colony (ABC) Algorithm demonstrate the better performance of the new algorithm.     

基于马氏距离的FCM图像分割算法

基于模糊C均值聚类的图像分割是应用较为广泛的方法之一,但大多数模糊C均值聚类方法都是基于欧式距离,且存在运算时间过长等问题。提出了一种基于Mahalanobis距离的模糊C均值聚类图像分割算法。实验分析表明,提出的算法在保证分割质量的前提下,能较快提高分割速度。实验结果表明了该方法的有效性。     

一种协同的可能性模糊聚类算法

模糊C-均值聚类（FCM）对噪声数据敏感和可能性C-均值聚类（PCM）对初始中心非常敏感易导致一致性聚类。协同聚类算法利用不同特征子集之间的协同关系并与其他算法相结合,可提高原有的聚类性能。对此,在可能性C-均值聚类算法（PCM）基础上将其与协同聚类算法相结合,提出一种协同的可能性C-均值模糊聚类算法（C-FCM）。该算法在改进的PCM的基础上,提高了对数据集的聚类效果。在对数据集Wine和Iris进行测试的结果表明,该方法优于PCM算法,说明该算法的有效性。     

基于L-ISOMAP降维的快速模糊聚类算法

模糊C-均值聚类算法是非监督模式识别中广泛应用的算法之一。但是,FCM算法在迭代过程中需要大量的计算,尤其当特征向量维数较高时,使用聚类分堆训练,不仅效率低下,还有可能导致“维数灾难”。针对该问题,分析模糊C-均值聚类算法在高维特征分析过程中,聚类中心的求解问题是一个np-hard问题,为了提高模糊C-均值聚类算法在高维特征分析中的实时性与有效性,结合界标等距映射（L-ISOMAP）算法,提出了改进算法FCM-LI,先对样本初步分析,利用聚类结果及样本数据相关性,使用界标等距映射（L-ISOMAP）算法降维,在此基础上进一步分析,获得最终分析结果。通过实验证明,FCM-LI算法在高维数据分析过程中的有效性与实时性。     

Study on multi-center fuzzy C-means algorithm based on transitive closure and spectral clustering

Fuzzy C-means (FCM) clustering has been widely used successfully in many real-world applications. However, the FCM algorithm is sensitive to the initial prototypes, and it cannot handle non-traditional curved clusters. In this paper, a multi-center fuzzy C-means algorithm based on transitive closure and spectral clustering (MFCM-TCSC) is provided. In this algorithm, the initial guesses of the locations of the cluster centers or the membership values are not necessary. Multi-centers are adopted to represent the non-spherical shape of clusters. Thus, the clustering algorithm with multi-center clusters can handle non-traditional curved clusters. The novel algorithm contains three phases. First, the dataset is partitioned into some subclusters by FCM algorithm with multi-centers. Then, the subclusters are merged by spectral clustering. Finally, based on these two clustering results, the final results are obtained. When merging subclusters, we adopt the lattice similarity method as the distance between two subclusters, which has explicit form when we use the fuzzy membership values of subclusters as the features. Experimental results on two artificial datasets, UCI dataset and real image segmentation show that the proposed method outperforms traditional FCM algorithm and spectral clustering obviously in efficiency and robustness.     

New fuzzy C-means clustering method based on feature-weight and cluster-weight learning

Among fuzzy clustering methods, fuzzy c-means (FCM) is the most recognized algorithm. In this algorithm, it is assumed that all the features are of equal importance. In real applications, however, the importance of the features are different and there exist some features that are more important than the others. These important features should basically have more effects than the other features in the forming of optimal clusters. The basic FCM algorithm does not support this idea. Also, the FCM algorithm suffers from another problem; the algorithm is very sensitive to initialization, whereas a bad initialization leads to a poor local optima. Some improved versions of FCM have been proposed in the literature, each of which has somehow mitigated the first problem or the second one. In this paper, motivated by these weaknesses of the FCM, the goal is to solve the two problems at the same time. In doing so, an automatic local feature weighting scheme is proposed to properly weight the features of each clusters. And, a cluster weighting process is performed to mitigate the initialization sensitivity of the FCM. Feature weighting and cluster weighting are performed simultaneously and automatically during the clustering process resulting in high quality clusters, regardless of the initial centers. Extensive experiments conducted on a synthetic dataset and 16 real world datasets indicate that the proposed algorithm outperforms the state-of-the-arts algorithms. The convergence proof of the proposed algorithm is also provided.