两阶段的多元时间序列异常检测算法*

提出了一个两阶段的多元时间序列异常检测算法。该算法通过有界坐标系统 （BCS）技术计算多元时间序列样本之间的相似性,采用基于距离的方法实现异常检测。算法第一阶段采用K-means算法对数据进行聚类,并按照一个启发式规则对其进行排序;第二阶段在聚类结果上采用循环嵌套算法进行异常检测,并通过两个剪枝规则进行高效剪枝,提高了算法的效率。在两个实际数据集上进行实验,实验结果验证了算法的有效性。     

一种支持快速相似检索的多维索引结构

基于内容的图像检索是一种典型的相似检索问题,对于尺度空间上的图像相似匹配问题,一般认为距离计算费用很高.因此,需要建立有效的索引结构,以减少每个查询中的距离计算次数.为此,基于数据空间的"优化划分",并且使用"代表点",以层次结构方式划分数据,提出了一种新的基于距离的相似索引结构opt-树及其变种(-树.为了更有效地支持基于内容的图像检索,在(-树索引结构中采用了"(-最优化划分"和"(-对称冗余存储"策略,以提高相似检索的效率.详细讨论了这种索引结构的建立与检索等问题,并给出了相应的算法.实验结果显示了这种索引技术的有效性.     

基于奇异值分解的异常切片挖掘

切片操作是联机分析处理的主要功能之一,在决策支持应用中发挥着重要作用.由于人工的切片过程非常低效,且易忽略重要信息,提出了一种自动、智能的异常切片挖掘方法.该方法基于奇异值分解技术来提取切片的数据分布特征,然后在提取出的奇异值特征之上,利用基于距离的孤立点检测技术发现异常的切片.在人工生成的数据和实际应用的切片数据上所作的实验结果都表明了该方法的高效性和可行性.     

离群点检测的邻近性方法综述

离群点检测在数据挖掘中有非常广泛的应用,然而并不是所有的离群点检测问题都能用一种最优的方法去解决。针对不同的应用,需要用不同的方法,才能够最有效地解决实际问题。检测方法大致可以分为基于统计、基于聚类、基于邻近性（基于距离和基于密度）的方法。为了及时掌握当前基于邻近性技术的离群点检测方法的研究现状,通过整理和归纳,将代表性强的基于邻近性的离群点检测方法进行了介绍和评价,将其主要分为基于距离的方法和基于密度的方法,对所有提及的方法的应用场景、算法思想、能解决的问题以及各自的优缺点进行了详细的分析和归纳,指出目前存在的问题和对未来研究的发展方向。对开展邻近性的离群点检测研究具有重要意义。     

Fuzzy Distance-Based Range Queries over Uncertain Moving Objects

Data obtained from real world are imprecise or uncertain due to the accuracy of positioning devices,updating protocols or characteristics of applications.On the other hand,users sometimes prefer to qualitatively express their requests with vague conditions and different parts of search region are in-equally important in some applications.We address the problem of efficiently processing the fuzzy range queries for uncertain moving objects whose whereabouts in time are not known exactly,for which the basic syntax is find objects always/sometimes near to the query issuer with the qualifying guarantees no less than a given threshold during a given temporal interval.We model the location uncertainty of moving objects on the utilization of probability density functions and describe the indeterminate boundary of query range with fuzzy set.We present the qualifying guarantee evaluation of objects,and propose pruning techniques based on the α-cut of fuzzy set to shrink the search space efficiently.We also design rules to reject non-qualifying objects and validate qualifying objects in order to avoid unnecessary costly numeric integrations in the refinement step.An extensive empirical study has been conducted to demonstrate the efficiency and effectiveness of algorithms under various experimental     

一种基于邻域系统密度差异度量的离群点检测算法

针对离群点检测算法LOF在高维离散分布数据集中检测精度较低及参数敏感性较高的问题,提出了基于邻域系统密度差异度量的离群点检测NSD（neighborhood system density difference）算法。相较于传统基于密度的离群点检测方法,NSD算法引入了截取距离的概念。首先计算数据集中对象在截取距离内的邻居点个数;其次计算对象的邻域系统密度;然后将对象的密度与它邻居的密度进行比较,判定目标对象与其邻居趋向于同一簇的程度;最后输出最可能是离群点的对象。将NSD算法与LOF、LDOF、CBOF算法在真实数据集与合成数据集中对比实验发现,NSD算法具有较高的检测准确率和执行效率以及较低的参数敏感性,证明了NSD算法是有效可行的。     

Research on the assessment of classroom teaching quality with q-rung orthopair fuzzy information based on multiparametric similarity measure and combinative distance-based assessment

The assessment of classroom teaching quality is critically important for producing a positive incentive and guidance role to improve service and management of universities, stimulating the enthusiasm of teachers, enhancing the teacher’s teaching ability, and improving the quality of talent training. In considering the case of teaching quality evaluation, the essential question that arises concerns strong ambiguity, fuzziness, and inexactness. The q-rung orthopair fuzzy sets ( q-ROFSs) dealing the indeterminacy characterized by membership degrees and nonmembership degrees are a more flexible and effective way to capture indeterminacy. In this paper, firstly, the new score function for q-rung orthopair fuzzy number is initiated for tackling the comparison problem. Subsequently, a new distance measure for q-ROFSs with multiple parameters is studied along with their detailed proofs. The various desirable properties among the developed similarity measures and distance measures have also been derived. Then, the objective weights of various attributes are determined via antientropy weighting method. Also, we develop the combined weights, which can reveal both the subjective information and the objective information. Moreover, two algorithms to solve q-rung orthopair fuzzy decision-making problem by combinative distance-based assessment and multiparametric similarity measure are presented. Later, the feasibility of approaches is demonstrated by a classroom teaching quality problem, along with the effect of the different parameters on the ordering. Finally, a comparison between the proposed and the existing decision-making methods has been performed for showing their effectiveness. The salient features of the proposed methods, compared to the existing q-ROFS decision-making methods, are as follows: (a) it can obtain the optimal alternative without counterintuitive phenomena and (b) it has a lower computational complexity.     

Agglomeration of creative industries: An intra-metropolitan analysis for Barcelona

In this paper we analyse the intensity and extent of agglomeration and co-agglomeration of creative industries (CIs) in the metropolitan area of Barcelona (MAB). To do so, we use firm-level geo-located data to calculate distance-based M and m functions of agglomeration and co-agglomeration. Our results show that CIs are relatively more agglomerated than non-CIs at short distances, and that each individual CI sector displays high levels of agglomeration (especially symbolic-based sectors). Also, the co-agglomeration of CIs and Non-CIs is only observed for micro-firms and that there is significant co-agglomeration among symbolic-based CI sectors. Finally, we confirm the role of the city-centre of Barcelona as a magnet for CIs.     

水下航行器声学故障检测中距离有效性的评价

引入J-偏差作为有向偏差的改进,并提出了一种新的距离有效性评价指标-归一化关键点高度(NKH).利用仿真数据和舱段试验数据,以线谱频率改变等4种类型的常见声学故障为研究对象,针对欧氏距离等5种距离定义,系统地开展声学故障检测有效性的评价研究.结果表明,J-偏差的检测效果要优于常用的欧式距离和Hausdorff距离.     

数据流环境下基于距离的离群点检测算法

面向滑动窗口的连续离群点检测问题是数据流管理领域中的重要问题.该问题在信用卡欺诈检测、网络入侵防御,地质灾害预警等诸多领域发挥着重要作用.现有算法大多需要利用范围查询判断对象之间的位置关系,而范围查询的查询代价大,无法满足实时性要求.本文提出基于滑动窗口模型下的查询处理框架GBEH(grid-based excepted heap).首先,它以网格为基础构建索引GQBI(grid queue based index)管理数据流.该索引一方面维护数据流之间的位置关系,另一方面利用队列维护数据流的时序关系.其次, GBEH提出离群点检测算法PBH(priority based heap).该算法利用查询范围与网格单元格的相交面积计算该单元格中包含于查询范围对象数目的数学期望,并以此为基础构建基于小顶堆执行范围查询,从而有效降低范围查询代价,实现高效检测.理论分析和实验验证GBEH的高效性和稳定性.