基于主题的用户兴趣域算法

针对用户兴趣偏好多变问题,提出一种兴趣特征权重随时间而变化的迭代计算方法。构造了用户兴趣特征与主题类间的二部图关系,并在此基础上提出了一种基于主题的用户兴趣聚类算法(TBC),改变了聚类对象"非此即彼"的硬划分方式。该算法所形成的基于主题的用户兴趣域结构,不仅充分表达了用户的多域兴趣特征和域间主题的联系,还能适应用户兴趣变化。实验表明,TBC算法比传统的K-Means算法以及属于软划分方式的FCM聚类具有更好的用户兴趣划分效果,并且在个性化推荐服务中表现出更高的推荐质量和效率。     

基于改进的密度峰值聚类的扩展目标跟踪算法

针对扩展目标高斯混合概率假设密度(extended target Gaussian mixture probability hypothesis density, ET-GM-PHD)滤波器中的量测集划分问题,提出了一种改进的密度峰值聚类(improved density peak clustering, IDPC)量测集划分算法。首先,使用IDPC算法去除局部密度较低的杂波量测,以获得最有可能的目标生成的量测集。其次,将剩余的量测集聚类以获得空间上紧密联系的聚类簇和簇的聚类中心。最后,根据预测的具有较高权重的高斯分量的均值在每个簇上的投影,获得准确的量测集划分。实验结果表明,与现有的量测集划分方法相比,该算法在保持跟踪精度的同时,可以大大减少计算时间。      

基于决策树的HEVC SCC帧内编码快速算法

为减少HEVC屏幕内容编码的编码时间,提高编码 效率,本文提出了一种基于决策树的HEVC屏幕内容帧内编码快速 CU划分和简单PU模式选择的算法。对视频序列特性分析,提取有效的特征值,生成决策树模 型。使用方差、梯度信息熵和 像素种类数用于生成CU划分决策树,使用平均非零梯度、像素信息熵等用于生成PU模式分类 决策树。在一定深度的决策 树模型中,通过对相应深度的CU的特征值的计算快速决策当前CU的划分与PU模式的类型。这 种利用决策树做判决的算法 通过减少CU深度和PU的模式遍历而降低编码复杂度,达到快速帧内编码的效果。实验结果表 明,与HEVC屏幕内容的标 准算法相比,该算法在峰值信噪比(PSNR)平均下降0. 05 dB和码率 平均增加1.15%的情况下,能平均减少30.81% 的编码时间。     

基于对称KL距离的用户行为时序聚类方法

网络用户随时间变化的行为分析是近年来用户行为分析的热点,通常为了发现用户行为的特征需要对用户做聚类处理。针对用户时序数据的聚类问题,现有研究方法存在计算性能差,距离度量不准确的缺点,无法处理大规模数据。为了解决上述问题,该文提出基于对称KL距离的用户行为时序聚类方法。首先将时序数据转化为概率模型,从划分聚类的角度出发,在距离度量中引入KL距离,用以衡量不同用户间的时间分布差异。针对实网数据中数据规模大的特点,该方法在聚类的各个环节针对KL距离的特点做了优化,并证明了一种高效率的聚类质心求解办法。实验结果证明,该算法相比采用欧式距离和DTW距离度量的聚类算法能提高4%的准确度,与采用medoids聚类质心的聚类算法相比计算时间少了一个量级。采用该算法对实网环境中获取的用户流量数据处理证明了该算法拥有可行的应用价值。     

基于信令数据的轨迹驻留点识别算法研究

针对密度聚类算法只能识别密度相近的簇类且计算复杂度高等问题,该文提出一种基于信令数据中时空轨迹信息的密度峰值快速聚类(ST-CFSFDP)算法。首先对低采样密度的信令数据进行预处理,消除轨迹震荡现象;然后基于密度峰值快速聚类(CFSFDP)算法显式地增加时间维度限制,将局部密度由2维扩展到3维,并提出高密度时间间隔以表征簇中心在时间维度上的数据特征;接着设计筛选策略以选取聚类中心;最后识别用户出行轨迹中的驻留点,完成出行链的划分。实验结果表明,所提算法适用于采样密度低且定位精度差的信令数据,相比CFSFDP算法更适用于时空数据,相比基于密度的时空聚类算法(ST-DBSCAN)召回率提升14%,准确率提升8%,同时降低计算复杂度。     

基于时间序列聚类的多雷达数据融合

针对多雷达数据融合问题，提出了基于时间序列的聚类算法，用于实现航迹相关，即以时间序列为基础把聚类模型转化为基于特征匹配的聚类算法。进一步考虑到多目标密集时，部分来自不同目标的数据可能比来自同一目标的数据更接近，易导致关联错误，为此提出了基于时间序列的模糊聚类算法。对上述两种算法的聚类结果，应用卡尔曼滤波器实现滤波跟踪，在不同的情况下仿真后发现，在跟踪目标较少且相互位置较远的情况下，两种算法均有效，在跟踪目标较多且相互位置靠近的情况下，基于时间序列的模糊聚类算法更有效。     

基于阴影集的多粒度三支聚类集成

聚类集成旨在通过融合多个不同的基聚类结果得到一个统一的类簇划分.针对现实环境中的模糊和不确定性数据,本文提出了一种基于阴影集的多粒度三支聚类集成算法.算法首先使用FCM聚类产生一组有差异性的基聚类成员,并通过阴影集构造三支聚类.然后引入多粒度粗糙集构建了四个近似集合,将每一个类簇划分为一个核心域和三个边界域.最后对边界域中的数据依次划分到核心域中,无法划分的对象则留在边界域,最终得到了三支聚类集成的结果.实验结果表明,本算法在准确率、调整兰德系数和归一化互信息方面,与多种现有的聚类集成算法相比得到了更好的聚类集成结果.     

面向MANET异常检测的分布式遗传k-means研究

针对移动自组网（MANET,mobile ad hoc networks）入侵检测过程中的攻击类型多样性和监测数据海量性问题,提出了一种基于改进k-means算法的MANET异常检测方法。通过引入划分贡献度的概念,可合理地计算各维特征在检测中占有的权重,并将遗传算法与快速聚类检测算法k-means相结合,解决了聚类检测结果容易陷入局部最优的问题,进而,提出了以上检测算法在MapReduce框架下的设计方案,利用种群迁移策略在分布式处理器上实现了并行聚类检测。实验结果证明了该方法的检测准确率和运行效率均优于传统聚类检测方法。     

改进粒子群结合K 均值聚类的图像分割算法

K 均值聚类的分类结果过分依赖于初始中心的选择且容易陷入局部最优。文中针对K 均值的缺陷,提出了一种基于随机权重粒子群和K 均值聚类的图像分割算法RWPSO KM。在算法开始,利用随机权重粒子群算法的全局搜索能力避免算法陷入局部最优。然后根据公式计算种群多样性执行K 均值算法,利用K 均值算法的局部搜索能力实现算法的快速收敛。实验结果表明, RWPSO KM与K 均值聚类和PSOK相比具有更好的分割效果和更高的分割效率。     

基于加权co-occurrence矩阵的聚类集成算法

聚类集成是数据挖掘研究的一个热点。它是利用同一数据集的多个聚类划分集成在一起,以提高聚类分析的性能。当前相关研究大多没有考虑进行集成的聚类成员的质量,因此较差的成员会对集成结果产生不良影响。文中提出了一种基于加权co-occurrence矩阵的聚类集成算法(WCSCE)。该方法首先计算出聚类成员基于属性值的co-occurrence矩阵,然后对聚类成员的质量进行简单评价并赋予权重,生成加权co-occurrence矩阵,进而产生集成结果。最后通过实验验证了该算法的有效性,并提高了聚类质量。     

New algorithms for learning and pruning oblique decision trees

We present methods for learning and pruning oblique decision trees. We propose a new function for evaluating different split rules at each node while growing the decision tree. Unlike the other evaluation functions currently used in the literature (which are all based on some notion of purity of a node), this new evaluation function is based on the concept of degree of linear separability. We adopt a correlation based optimization technique called the Alopex algorithm (K.P. Unnikrishnaan and K.P. Venugopal, 1994) for finding the split rule that optimizes our evaluation function at each node. The algorithm we present is applicable only for 2-class problems. Through empirical studies, we demonstrate that our algorithm learns good compact decision trees. We suggest a representation scheme for oblique decision trees that makes explicit the fact that an oblique decision tree represents each class as a union of convex sets bounded by hyperplanes in the feature space. Using this representation, we present a new pruning technique. Unlike other pruning techniques, which generally replace heuristically selected subtrees of the original tree by leaves, our method can radically restructure the decision tree. Through empirical investigation, we demonstrate the effectiveness of our method     

C-fuzzy decision trees

This paper introduces a concept and design of decision trees based on information granules - multivariable entities characterized by high homogeneity (low variability). As such granules are developed via fuzzy clustering and play a pivotal role in the growth of the decision trees, they will be referred to as C-fuzzy decision trees. In contrast with "standard" decision trees in which one variable (feature) is considered at a time, this form of decision trees involves all variables that are considered at each node of the tree. Obviously, this gives rise to a completely new geometry of the partition of the feature space that is quite different from the guillotine cuts implemented by standard decision trees. The growth of the C-decision tree is realized by expanding a node of tree characterized by the highest variability of the information granule residing there. This paper shows how the tree is grown depending on some additional node expansion criteria such as cardinality (number of data) at a given node and a level of structural dependencies (structurability) of data existing there. A series of experiments is reported using both synthetic and machine learning data sets. The results are compared with those produced by the "standard" version of the decision tree (namely, C4.5).     

基于二叉树锚点的高光谱快速聚类算法

高光谱图像聚类问题一直是图像处理领域的研究热点。谱聚类算法是最流行的聚类算法之一,但其计算复杂度较大,难以处理大规模的高光谱图像数据。由于二叉树能够较快地选取锚点,因此基于二叉树锚点图,充分利用高光谱图像的光谱和空间特性,可保证聚类性能并降低计算复杂度。然而,该聚类算法一般采用有核的聚类方法,因此不可避免地引入了参数调节。在二叉树锚点选取的基础上,提出了一种基于二叉树锚点的高光谱快速聚类算法,该算法创新性地将二叉树锚点选取和无核聚类方法应用于高光谱图像中。首先,利用二叉树从高光谱数据中选取一些具有代表性的锚点;紧接着构造基于锚点的无核相似图,有效避免了通过人为调节热核参数来构造相似图;然后进行谱聚类分析获得聚类结果;最后,将该算法应用到高光谱图像聚类中。该算法不仅提高了聚类速度,还减少了原有热核参数调节。实验结果表明,与传统的聚类算法相比,所提算法能够在较短的时间内获得更佳的聚类精度。     

KFDA and clustering based multiclass SVM for intrusion detection

To improve the classification accuracy and reduce the training time, an intrusion detection technology is proposed, which combines feature extraction technology and multiclass support vector machine （SVM） classification algorithm. The intrusion detection model setup has two phases. The first phase is to project the original training data into kernel fisher discriminant analysis （KFDA） space. The second phase is to use fuzzy clustering technology to cluster the projected data and construct the decision tree, based on the clustering results. The overall detection model is set up based on the decision tree. Results of the experiment using knowledge discovery and data mining （KDD） from 99 datasets demonstrate that the proposed technology can be an an effective way for intrusion detection.     

Minimizing Cost and Delay in Shared Multicast Trees

Existing tree construction mechanisms are classified into source‐based trees and center‐based trees. The source‐based trees produce a source‐rooted tree with a low delay. However, for the applications with multiple senders, the management overheads for routing tables and resource reservations are too high. The center‐based trees are easy to implement and manage, but a priori configuration of candidate center nodes is required, and the optimization nature such as tree cost and delay is not considered. In this paper, we propose a new multicast tree building algorithm. The proposed algorithm basically builds a non‐center based shared tree. In particular, any center node is not pre‐configured. In the proposed algorithm, a multicast node among current tree nodes is suitably assigned to each incoming user. Such a node is selected in a fashion that tree cost and the maximum end‐to‐end delay on the tree are jointly minimized. The existing and proposed algorithms are compared by experiments. In the simulation results, it is shown that the proposed algorithm approximately provides the cost saving of 30 % and the delay saving of 10 %, compared to the existing approaches. In conclusion, we see that the cost and delay aspects for multicast trees can be improved at the cost of additional computations.     

Multicriteria Steiner tree with the cost of Steiner vertices

The multicriteria problem of constructing the Steiner tree with consideration for the cost of additional Steiner vertices is studied. Formulations of the problems of constructing the covering Steiner trees and algorithmic approaches are described. The engineering formulation of the problem is aimed at constructing a telecommunications network with allowance for the spatial distribution of radio interferences. An approach to solving the formulated milticriterion problem on the basis of combination of two solving schemes is proposed: (1) the basic scheme for constructing a covering Steiner tree on the basis of clustering the vertices of the initial graph and (2) the general scheme for constructing the covering Steiner trees with allowance for the number of additional vertices and calculation of the Pareto-efficient solutions. A module based on the modified Prim’s algorithm for constructing a multicriteria covering tree is additionally used. The vertices of the initial graph are clustered on the basis of a hierarchical algorithm. Examples of a computational experiment on constructing the multicriteria Steiner trees are given.     

基于Spark的分层子空间权重树随机森林算法

高维数据的很多特征与类别的相关性弱,影响了随机森林的分类正确率。针对原始随机森林算法在高维数据上的分类问题,提出了一种分层子空间权重树随机森林算法。同时,传统的单机模式无法满足高维数据计算效率的需求,因此利用开源集群计算框架Spark在内存缓存和迭代计算上的优势,将所提算法在Spark上实现。所提算法采用以决策树为单位的分层抽样来生成特征子空间,在提高单棵决策树性能的同时,保证决策树之间的多样性;并且采用权重树的集成策略,使分类能力强的树在集成过程中影响力更大。通过在Mnist和Gisette数据集上的实验结果表明,相比原始随机森林算法、TWRF算法以及分层子空间随机森林算法,所提算法具有更好的正确率,提高了泛化误差性能,可扩展性良好,能够有效分类高维数据。      

一种基于NNIA多目标优化的代价敏感决策树构建方法

本文提出了一种基于非支配邻域免疫算法(NNIA,Nondominated Neighbor Immune Algorithm)多目标优化的代价敏感决策树构建方法.将平均误分类代价和平均测试代价作为两个优化目标,然后利用NNIA对决策树进行优化,最终获取了一组Pareto最优的决策树.对多个测试集的测试结果表明,与C4....     

A greedy tree growing algorithm for the design of variable ratevector quantizers [image compression]

A technique for directly designing a variable-rate tree-structured vector quantizer by growing the tree one node at a time rather than one layer at time is presented. The technique is a natural extension of a tree growing method for decision trees. When the tree is pruned with a generalized algorithm for optimally pruning trees, improvement is measured in the signal-to-noise ratio at high rates over pruning a fixed-rate tree-structured vector quantizer of the same initial rate. The growing algorithm can be interpreted as a constrained inverse of the pruning algorithm