基于投影和密度的高维数据流聚类算法

在经典数据流的聚类算法基础之上,提出了一种基于投影和密度的高维数据流聚类算法——HpDenStream,该算法结合滑动窗口技术,采用投影算法对高维数据流进行降维处理,并运用密度聚类算法对降维后的数据进行异常数据检测。仿真实验结果表明:该方法占用的存储空间小,算法的工作量少,并提高了算法的执行效率。     

基于决策属性的关联规则挖掘

针对具有决策属性的数据库模型,提出了高效挖掘关联规则算法,即矩阵划分算法:根据决策属性将扫描后的数据库划分成两个包含不同决策属性的矩阵,分别采用向量法挖掘频繁项目集.关联规则的生成可充分利用“与”运算的优点,查找规则前件或后件的支持度.所提出的算法减少了候选二项频集的生成,以及“与”运算的大小,与apriori算法及传统的向量法挖掘关联规则相比,效率明显提高.     

基于决策属性的关联规则挖掘

针对具有决策属性的数据库模型,提出了高效挖掘关联规则算法,即矩阵划分算法:根据决策属性将扫描后的数据库划分成两个包含不同决策属性的矩阵,分别采用向量法挖掘频繁项目集.关联规则的生成可充分利用"与"运算的优点,查找规则前件或后件的支持度.所提出的算法减少了候选二项频集的生成,以及"与"运算的大小,与apriori算法及传统的向量法挖掘关联规则相比,效率明显提高.     

基于异几率属性的可视化关联规则挖掘

在分析数据特性的基础上,提出了一种基于异几率属性的可视化关联规则挖掘算法,不仅提高了原算法的运行效率,而且提供了一个可视化的交互平台,使用户主动地挖掘感兴趣的关联规则.     

一种小支持度布尔关联规则的挖掘方法

关联规则挖掘算法一般用于发现强关联规则,对于小支持度规则的挖掘则缺少有效的算法.利用事务数据的时间特性,将事务数据集划分成若干子集,对子集进行挖掘,并在得到的规则集基础上建立规则矩阵,过滤矩阵,得到一种挖掘事务数据集中小支持度布尔关联规则的新方法.     

基于个体出行数据的公共交通出行链提取方法

公共交通个体出行信息的提取对掌握公共交通出行的时空特征,改善居民通勤出行效率具有重要意义.研究从公交刷卡数据、公交定位数据、轨道AFC数据等海量公共交通多源数据的关联匹配与处理方法入手,提出了公共交通出行链信息提取中,换乘关系判断、通勤行为判别及出行起讫点匹配的方法与规则,标定了出行链匹配阈值参数,建立了基于个体出行数据的公共交通通勤出行链提取模型.提取模型的准确度验证表明:出行链结构提取及通勤出行判别的成功率均达到100%,出行阶段起讫点匹配成功率为87.5%,准确性为97.1%,满足了公共交通出行特征提取的需求.该方法为公共交通通勤出行判别及基于个体的微观通勤出行时空特征的深入分析奠定了基础.     

基于粗约简的数据流增量聚类算法

针对数据流聚类算法CluStream需预先指定微聚类数目无法准确描述数据流的变化,进而影响最终聚类结果的缺陷,提出了基于粗约简的数据流增量聚类算法RICStream(rough incremental clustering stream).该算法在保证聚类精度的前提下,对参与聚类的数据流属性进行动态调整,有效地减少了聚类时间和计算量.提出了一种可增量调整的网格结构以存储数据流,保证了聚类结果能有效反映数据流的变化情况.基于真实数据集和仿真数据集的实验结果表明,RICStream算法具有较高的效率和聚类精度.     

基于点特征的图像数据配准方法研究

数据配准是数据融合过程中最重要的一个环节,通过感兴趣点检测算子提取点特征,利用点特征中的控制点或特殊点来定义松弛算法中的基本点对,并对松弛算法中点特征的匹配度进行改进,从而完成基于点特征的图像配准过程.实验表明,在仅存点特征位置偏移和存在比例、旋转与平移变化2种情况下的点特征匹配中,文中所提出的匹配算法有很高的匹配正确率.     

基于异几率属性的可视化关联规则挖掘

在分析数据特性的基础上，提出了一种基于异几率属性的可视化关联规则挖掘算法，不仅提高了质算法的运行效率，而且提供了一个可视化的交互平台，使用户主动地挖掘感兴趣的关联规则。     

基于语义相似度的话题关联检测方法

为有效识别任意两篇报道的相似性,提出了一种基于语义相似度的话题关联检测算法.该算法首先通过计算特征词之间的相对熵作为两篇报道中特征词之间的语义相似度;其次,通过计算平均语义相似度获得特征词和报道之间的关联度;最后,结合特征词在语料库中的TF-IF(term frequency-inverse document frequency)权重计算两篇报道之间的关联度,实现报道之间的关联度检测.本文提出的方法与现有的向量空间模型方法和仅依赖于平均点互信息的方法进行了比较,并通过TDT4中文语料进行测评,结果表明,基于语义相似度的关联检测方法能够更好地利用文本的语境信息,提高了现有检测系统的性能,其最小DET(detection error tradeoff)代价降低了3%.      

A Robust Interacting Multisensor State Fusion Based on Adaptive Outlier Controlling Multirate Model

An adaptive outlier controlling multirate model based on Hong’s multirate kinetic model was represented in order to resist the outliers and utilize their useful information. Wavelet transform was introduced to detect and control the outliers. The multirate information extraction and the controlling of outliers were properly integrated to establish an adaptive outlier controlling multirate model. The proposed model was applied to multisensor state fusion with interacting multiple model (IMM), and a robust interacting multisensor state fusion algorithm was established based on adaptive outlier controlling multirate model. The Monte-Carlo simulation shows that it could improve the accuracy of fusion estimation by 70% compared to Hong’s algorithm and at least 14% to Xiao’s algorithm.     

Mining φ-Frequent Itemset Using FP-Tree

The problem of association rule mining has gained considerableprominence in the data mining community for its use as an important tool of knowledge discovery from large-scale databases. And there has been a spurt of research activities around this problem. However, traditional association rule mining may often derive many rules in which people are uninterested. This paper reports a generalization of association rule mining called φ-association rule mining. It allows people to have different interests on different itemsets that are the need of real application. Also, it can help to derive interesting rules and substantially reduce the amount of rules. An algorithm based on FP-tree for mining φ-frequent itemset is presented. It is shown by experiments that the proposed method is efficient and scalable over large databases.     

Outlier Rejecting Multirate Model for State Estimation

IntroductionMeasured data is often contaminated by noisein state estimation.Kalman filter is a powerfultool for signal extracting.It is especially efficientin estimating spatially inhomogeneous signal whenthe noise is Gaussian.Due to process noise or non-stationary environment,the measured data is usu-ally corrupted by outliers.The performance is de-graded seriously.Generally,there are two kinds ofapproaches to handle this problem.Outlier can bedetected based on renovation[1],then be replace…     

Outlier Rejecting Muitirate Model for State Estimation

Wavelet transform was introduced to detect and eliminate outliers in time-frequency domain. The outlier rejection and multirate information extraction were initially incorporated by wavelet transform, a new outlier rejecting multirate model for state estimation was proposed. The model is applied to state estimation with interacting multiple model, as the outlier is eliminated and more reasonable multirate information is extracted, the estimation accuracy is greatly enhanced. The simulation results prove that the new model is robust to outliers and the estimation performance is significantly improved.     

Outlier mining based abnormal machine detection in intelligent maintenance

Assessing machine’s performance through comparing the same or similar machines is important to implement intelligent maintenance for swarm machine. In this paper, an outlier mining based abnormal machine detection algorithm is proposed for this purpose. Firstly, the outlier mining based on clustering is introduced and the definition of cluster-based global outlier factor (CBGOF) is presented. Then the modified swarm intelligence clustering(MSIC) algorithm is suggested and the outlier mining algorithm based on MSIC is proposed. The algorithm can not only cluster machines according to their performance but also detect possible abnormal machines. Finally, a comparison of mobile soccer robots’ performance proves the algorithm is feasible and effective.     

求解高次方程的一个异步并行迭代算法

用高次方程正项分解方法，将求解实系数高次方程非零实数根的问题，转化成求解两单调上升凹函数在平面直角系第一象限内交点横坐标的等价问题；给出了基于共享存储多指令流多数据流(MIMD)并行计算模型求解任意实系数高次方程全部实数根的大范围收敛性异步并行迭代算法，并分析了算法计算的复杂程度。     

Logistic Regression for Evolving Data Streams Classification

Logistic regression is a fast classifier and can achieve higher accuracy on small training data.Moreover,it can work on both discrete and continuous attributes with nonlinear patterns.Based on these properties of logistic regression,this paper proposed an algorithm,called evolutionary logistical regression classifier(ELRClass),to solve the classification of evolving data streams.This algorithm applies logistic regression repeatedly to a sliding window of samples in order to update the existing classifier,to keep this classifier if its performance is deteriorated by the reason of bursting noise,or to construct a new classifier if a major concept drift is detected.The intensive experimental results demonstrate the effectiveness of this algorithm.     

A two-step outlier removal procedure for surface reconstruction

Outliers in point clouds affect the performance of surface reconstruction directly. Most of outlier removal methods just remove those outliers far away from the real surface and are only applied to handle watertight surface. In this paper, a two-step outlier removal procedure is proposed to filter the point clouds acquired from the gray code and line-shifting technique. The first step is to remove the outliers far away from the real surface. Some feature points are extracted from the point clouds to construct an initial surface. The points with distances to the initial surface greater than a given threshold are removed as distant outliers. The retained points are linked into lines in each structured light sheet using their Voronoi diagrams. Some of lines which are very close to the real surface are removed as near outliers in the second step. The experimental results show that the proposed method is very effective in removing outliers for surface reconstruction. Foundation item: the National Natural Science Foundation of China (No. 30470488)     

BM算法的研究与改进

随着网络的迅速发展，网络安全问题日益突出，入侵检测技术也成为当今社会关注的焦点.对于基于规则的入侵检测来说，模式匹配算法非常重要，它直接影响到系统的准确性和实时性能.文中介绍了KMP和BM算法，对BM算法的改进进行了研究，并提出一种改进的BM算法，改进后的算法极大地提高了匹配速度．     

Fast triangulation and local optimization for scan data of laser radar

In order to study the triangulation for the point cloud data collected by three-dimensional laser radar, in accordance with the line-by-line characteristics of laser radar scanning, an improved Delaunay triangulation method is proposed to mesh the point cloud data as a triangulation irregular network. Based on the geometric topology location information among radar point cloud data, focusing on the position relationship between adjacent scanning line of the point data, a preliminary match network is obtained according to their geometric relationship. A reasonable triangulation network for the object surface is acquired after the use of local optimization on initial mesh by Delaunay rule. Meanwhile, a new judging rule is proposed to contrast the triangulation before and after the optimization on the network. The result shows that triangulation for point cloud with full use of its own characteristics can improve the speed of the algorithm obviously, and the rule for judging the triangulation can evaluate the quality of network.