数据流滑动窗口方式下的自适应集成分类算法

针对基于数据块的集成算法,存在数据块大小影响分类效果,且不能及时应对完整式概念漂移的问题,提出了一种考虑数据流局部特征的和能应对多种类型概念漂移的集成分类算法.用滑动窗口作为概念漂移检测器,当检测到概念漂移时,则建立新的分类器并加入到集成分类器中.本文提出的算法在人工合成和真实数据集上与经典算法进行了广泛的对比实验.结果表明:提出的算法在分类准确率上具有明显优势,消耗更少的内存,更适合多种类型概念漂移的环境.     

Linked-Tree： An Aggregate Query Algorithm Based on Sliding Window over Data Stream

How to process aggregate queries over data streams efficiently and effectively have been becoming hot re search topics in both academic community and industrial community. Aiming at the issues, a novel Linked-tree algorithm based on sliding window is proposed in this paper. Due to the proposal of concept area, the Linked-tree algorithm reuses many primary results in last window and then avoids lots of unnecessary repeated comparison operations between two successive windows. As a result, execution efficiency of MAX query is improved dramatically. In addition, since the size of memory is relevant to the number of areas but irrelevant to the size of sliding window, memory is economized greatly. The extensive experimental results show that the performance of Linked-tree algorithm has significant improvement gains over the traditional SC （Simple Compared） algorithm and Ranked-tree algorithm.     

A graph based sliding window multi join over data stream

Join operation is a critical problem when dealing with sliding window over data streams. There have been many optimization strategies for sliding window join in the literature, but a simple heuristic is always used for selecting the join sequence of many sliding windows, which is ineffectively. The graphbased approach is proposed to process the problem. The sliding window join model is introduced primarily. In this model vertex represent join operator and edge indicated the join relationship among sliding windows. Vertex weight and edge weight represent the cost of join and the reciprocity of join operators respectively. Then good query plan with minimal cost can be found in the model. Thus a complete join algorithm combining setting up model, finding optimal query plan and executing query plan is shown. Experiments show that the graphbased approach is feasible and can work better in above environment     

A graph-based sliding window multi-join over data stream

Join operation is a critical problem when dealing with sliding window over data streams. There have been many optimization strategies for sliding window join in the literature, but a simple heuristic is always used for selecting the join sequence of many sliding windows, which is ineffectively. The graph-based approach is proposed to process the problem. The sliding window join model is introduced primarily. In this model vertex represent join operator and edge indicated the join relationship among sliding windows. Vertex weight and edge weight represent the cost of join and the reciprocity of join operators respectively. Then good query plan with minimal cost can be found in the model. Thus a complete join algorithm combining setting up model, finding optimal query plan and executing query plan is shown. Experiments show that the graph-based approach is feasible and can work better in above environment.     

Dynamically Computing Approximate Frequency Counts in Sliding Window over Data Stream

This paper presents two one-pass algorithms for dynamically computing frequency counts in sliding window over a data stream-computing frequency counts exceeding user-specified threshold ε. The first algorithm constructs subwindows and deletes expired sub-windows periodically in sliding window, and each sub-window maintains a summary data structure. The first algorithm outputs at most 1/ε ＋ 1 elements for frequency queries over the most recent N elements. The second algorithm adapts multiple levels method to deal with data stream. Once the sketch of the most recent N elements has been constructed, the second algorithm can provides the answers to the frequency queries over the most recent n （ n≤N） elements. The second algorithm outputs at most 1/ε ＋ 2 elements. The analytical and experimental results show that our algorithms are accurate and effective.     

滑动窗口内动态数据流聚类算法研究

滑动窗口是数据流中一种关注近期数据的近似方法,提出一种采用滑动窗口处理数据的优化算法SWStream。在线阶段利用滑动窗口树存储概要结构,动态调整窗口大小。优化后的算法能及时淘汰过期元组,同时对新到达的元组不断进行实时处理,可以获得更准确的分析结果。而在离线阶段对上一阶段的结果进行宏聚类,得到最后的结果。与聚类算法CluStream相比,此算法处理数据的效率更高,也相对节约内存。     

数据流中一种有效的当前频繁序列挖掘方法

给出了一种基于滑动窗口挖掘频繁序列算法。该算法给出了ε-近似序列集的定义,利用一种压缩的数据结构GSP-tree来存储和维护整个滑动窗口中各分区的近似序列集,并通过合并各分区的近似序列集来响应用户当前的查询请求。     

数据流滑动窗口连接的自适应降载策略

连接操作是一种应用广泛的查询类型,在对DSMS进行降载时,必须考虑连接查询的降载问题.针对滑动窗口连接查询,构建一种基于不同等级基本窗口的滑动窗口内部结构.采用部分连接的方式实现CPU过载情况下的降载,采用基于基本窗口重要性的语义丢弃方式实现内存不足情况下的降载.实验结果表明所采用的算法有效.     

复合滑动窗口连接算法

滑动窗口是对数据流进行连续查询处理、得出近似查询结果的一种常用的数据采样技术．连续查询处理的执行方式有2类：一类是立即执行方式；一类是周期执行方式、在已有的研究工作中，基于滑动窗口的连接算法都是适用于立即执行的连续查询．提出了适用于周期执行的连续查询的复合滑动窗口连接算法，即CSWSNLJ算法和CSWSNHJ算法．理论分析和试验结果表明CSWSNHJ算法具有较好的性能．     

数据流时间窗口中闭频繁项集的在线挖掘

在数据流闭频繁项集挖掘过程中,常忽略历史模式对挖掘结果的影响,并采用一种结构来标记闭频繁项集的类型,导致算法的效率不高.为此提出一种挖掘数据流时间窗口中闭频繁项集的方法NEWT-moment.该方法能在单遍扫描数据流事务的条件下完整地记录模式信息.同时,NEWT-moment提出的剪枝方法能很好地降低滑动窗口树F-tr...     

一种基于预裁剪的不确定数据流频繁项集挖掘算法

随着交通、网络流量监控等应用的涌现,不确定数据流频繁项集挖掘成为近年来的研究热点。通常在不确定数据流中,频繁项集所占的比例较小,导致挖掘中无效操作较多。基于这种情况,提出了一种基于预裁剪的不确定数据流频繁项集挖掘算法Prep-UF-Streaming;该算法,不仅能裁剪掉大部分非频繁项集,提高了算法的平均运行时间;而且能够检测到非频繁项集成为频繁项集的可能性,尽量不丢失频繁项集,从而尽可能地提高算法的性能。     

基于近似等深柱状图的数据流并行聚集算法

针对数据流并行聚集问题,提出了一种不同于关系数据和时间序列数据处理的并行聚集方法.为解决已经划分出的数据流元组无法再现的特点,提出能够感知数据流变化的采样算法对数据流采样.利用近似等深柱状图技术描述采样数据的分布特征,平均分配数据流量.使用时间聚集森林结构计算时间窗聚集.通过验证采样个数对并行聚集的影响,数据分布对近似划分向量算法性能的影响,测试数据流量与并行聚集加速比的关系,证明本算法能够高效地计算数据流聚集查询.     

基于MapReduce的数据流频繁项集挖掘算法

针对传统数据流频繁项集计算中效率低、内存消耗大等问题,本文采用并行计算的思想设计了一种基于MapReduce的数据流频繁项集挖掘算法,首先,对进行数据分块压缩和传输,其次,将数据频繁项的计算分布在负载均衡的数据节点,可以有效保证数据的执行效率.最后通过一次调度处理合并各个节点产生的频繁项集并进行合并.理论分析和实验对比结果均表明,该算法对于并行处理数据流频繁项集的统计问题是有效可行的.     

基于随机平衡采样的不平衡数据流分类研究

数据流广泛应用于现实世界的多个领域,但是不平衡数据流的存在严重影响了传统数据流分类器的性能.针对不平衡数据流问题,提出了随机平衡采样算法(RBS)处理数据流的不平衡问题,并以RBS算法为基础提出了随机平衡采样数据流集成算法(RBSSEA)旨在解决不平衡数据流的分类问题.最后,分别采用合成和真实数据集对RBSSEA算法进行验证,实验结果证明RBSSEA算法在解决不平衡数据流分类问题具有一定的优势.     

基于ASWT的流数据弹性跳变检测算法

在迁移小波树的基础上提出了带放大器的迁移小波树(ASWT)结构,并基于该结构设计出了一个能够在线性时间内检测出弹性窗口中感兴趣的聚集的流数据弹性跳变检测算法.该算法不仅能够监测单调聚集(COUNT、SUM、MAX和MIN)跳变,也适应非单调聚集(AVG)跳变检测.输入数据流不受限制,可以是正数流、负数流或者正负数交错流.实验使用上海证券交易所的实时贸易数据监测交易高峰期和股票价格的涨停期,结果表明本算法是较高效的.     

基于非平衡数据的随机森林分类算法改进

随机森林算法作为一种组合分类器有较好的分类性能,适合多样的分类环境。算法同样也存在一些缺陷,例如算法处理非平衡数据时不能很好地区分正类和负类。针对这一问题,通过对抽样结果增加约束条件来改进Bootstrap重抽样方法,减少抽样对非平衡性的影响,同时尽量保证算法的随机性。之后利用生成数据的非平衡系数给每棵决策树进行加权处理,提升对非平衡数据敏感的决策树在投票环节的话语权,从而提升整体算法对非平衡数据的分类性能。通过上述两种改进可以明显提高随机森林在决策树数量不足情况下的分类精度。     

基于时间窗口权值的数据流分类算法

针对不同时间段的数据流对当前数据流分类影响程度不同,在滑动窗口技术基础上提出了时间窗口权值的频繁模式(TWWFP)分类算法.首先,对滑动窗口中的每个基本窗口赋予一个与时间有关的窗口权值;然后,采用TWWFP-Tree结构存储当前滑动窗口中每个基本窗口中的频繁数据属性,实时更新TWWFP-Tree结构;最后,检测相邻3个滑动窗口中权值属性的平均分类误差,发现突变后及时减少下一个滑动窗口的长度可适应数据流的变化.实验证明该分类算法比没有时间窗口权值分类算法的精确度最大提高3%.     

Path planning based on sliding window and variant A*algorithm for quadruped robot

In order to improve the adaptability of the quadruped robot in complex environments , a path planning method based on sliding window and variant A * algorithm for quadruped robot is presen-ted .To improve the path planning efficiency and robot security , an incremental A*search algorithm ( IA*) and the A*algorithm having obstacle grids extending ( EA*) are proposed respectively .The IA* algorithm firstly searches an optimal path based on A * algorithm, then a new route from the current path to the new goal projection is added to generate a suboptimum route incrementally .In comparison with traditional method solving path planning problem from scratch , the IA* enables the robot to plan path more efficiently .EA* extends the obstacle by means of increasing grid g-value, which makes the route far away from the obstacle and avoids blocking the narrow passage .To navi-gate the robot running smoothly , a quadratic B-spline interpolation is applied to smooth the path . Simulation results illustrate that the IA* algorithm can increase the re-planning efficiency more than 5 times and demonstrate the effectiveness of the EA * algorithm.     

基于时间滑动窗口的自适应加权随机抽样算法

为了构建传感器网络流数据的概要数据,给出了一种基于时间滑动窗口的自适应加权随机抽样算法:AWRS/BTSW算法.算法根据流数据的到达时间和变化情况,赋予流数据一定的键值,根据流数据的键值并结合skipping因子生成概要数据.在深海平台监测系统中,流数据变化不确定,算法可以根据数据的变化情况动态调整抽样方式,在数据变化不稳定的情况下,生成概要数据的准确性高;在数据变化稳定的情况下,生成概要数据的效率高.     

A nearest neighbor search algorithm of high-dimensional data based on sequential NPsim matrix

Problems existin similarity measurement and index tree construction which affect the perform-ance of nearest neighbor search of high-dimensional data .The equidistance problem is solved using NPsim function to calculate similarity .And a sequential NPsim matrix is built to improve indexing performance .To sum up the above innovations , a nearest neighbor search algorithm of high-dimen-sional data based on sequential NPsim matrix is proposed in comparison with the nearest neighbor search algorithms based on KD-tree or SR-tree on Munsell spectral data set .Experimental results show that the proposed algorithm similarity is better than that of other algorithms and searching speed is more than thousands times of others .In addition , the slow construction speed of sequential NPsim matrix can be increased by using parallel computing .