基于隐马尔可夫模型的兴趣迁移模式发现

Web挖掘的一个重要研究方向是发现用户的迁移模式。一般来说,用户的迁移具有某种目的性。这种目的性表现为用户对某种概念的兴趣。文中提出基于隐马尔可夫模型的兴趣迁移模式发现方法,用于发现这种带有某种兴趣的用户迁移模式,这种模式实质上是一种特殊的关联规则。在这种方法中,作者首先根据用户的访问记录定义一个隐马尔可夫模型,然后提出一种新的增量发现算法Increase_R用于发现兴趣迁移模式,同时给出了证明以说明该算法可以发现所有的兴趣迁移模式。     

基于混合行为兴趣度的用户兴趣模型

用户兴趣模型是个性化推荐技术的基础与核心,针对现有用户兴趣模型在模型建立阶段用户兴趣评价的不足,提出了一种基于混合行为评价兴趣度值的方法,突出了用户阅读时间的特殊性,在用户阅读时间异常的情况下利用其它浏览行为来量化用户兴趣度,并结合用户的浏览内容提出了用户兴趣模型的表示和更新机制,从而建立用户兴趣模型。实验验证了兴趣度度量方法的有效性,将测试结果与K-Means聚类模型进行比较,该模型的推荐准确度有明显提高。     

基于隐马尔可夫模型的躯感网心电图信号特征提取方法

为了解决躯感网的心电信号特征提取问题,结合心电图信号波形的特征区间,建立了面向心电图信号特征提取的离散隐马尔可夫模型;并面向该模型定制了专家标注选取、导联选取、观察数据归一化、三元组初始值选取以及训练数据量选取等方法.最后,采用Baum-Welch算法训练HMM模型的参数,并利用Viterbi算法提取心电图的信号特征.仿真结果表明,基于 HMM的心电图信号特征提取算法的复杂度较低、精确度较高、实时性较好,适合在线处理非线性、动态变化的心电图信号,能够满足基于躯感网的心电图信号特征提取的性能要求.     

基于隐马尔可夫模型的动态手势识别研究

随着虚拟现实技术的飞速发展，人们迫切需要一种自然友好的字符输入方式，于是越来越多的研究人员投入到动态手势的研发当中。本文基于隐马尔可夫模型（HMM）搭建了一套动态手势识别系统。这套系统通过Leap Motion采集动态手势数据，并能够识别36个字母和数字的手势（数字0-9和字母A-Z）。经过大量实验表明，该系统有着很强的鲁棒性，识别单独手势的识别率能够达到93.2%。     

基于隐式反馈的自适应用户兴趣模型

针对现有用户兴趣模型在模型建立以及更新阶段漂移策略的缺陷,设计了一种改进的基于隐式反馈的自适应用户兴趣模型。并将该模型与遗忘策略模型、滑动时间窗口模型和固定比例模型,做了精确率的对比实验,实验结果显示该模型的性能优于其他三个模型。     

基于农业本体的用户兴趣模型研究

为了有效解决传统用户兴趣模型查不全,查不准等问题,引入农业本体技术构建用户兴趣模型。该模型能在语义层次上理解用户的兴趣,因而在检索时能获取较满意的查全率和查准率,能更好的体现农户的个性化需求。     

基于微博用户兴趣话题的相似用户挖掘

相似用户挖掘是提高社交网络服务质量的重要途径，在面向大数据的社交网络时代，准确的相似用户挖掘对于用户和互联网企业等都有重要的意义，而根据用户自己的兴趣话题挖掘的相似用户更符合相似用户的要求。提出了一种基于用户兴趣话题进行相似用户挖掘的方法。该方法首先使用TextRank话题提取方法对用户进行兴趣话题提取，再对用户发表内容进行训练，计算出所有词之间的相似度。提出CP（Corresponding Position similarity）、CPW（Corresponding Position Weighted similarity）、AP（All Position similarity）、APW（All Position Weighted similarity）四种用户兴趣话题词相似度计算方法，通过用户和相似用户间关注、粉丝重合率验证相似用户挖掘效果，APW similarity的相似用户的关注/粉丝重合百分比为1.687%，优于提出的其他三种算法，分别提高了26.3%、2.8%、12.4%，并且比传统的文本相似度方法Jaccard相似度、编辑距离算法、余弦相似度分别提高了20.4%、21.2%、45.0%。因此APW方法可以更加有效地挖掘出用户的相似用户。     

基于隐半马尔可夫模型的SWIM应用层DDoS攻击的检测方法

针对广域信息管理（SWIM）系统受到应用层分布式拒绝服务（DDoS）攻击的问题，提出了一种基于隐半马尔可夫模型（HSMM）的SWIM应用层DDoS攻击的检测方法。首先采用改进后的前向后向算法，利用HSMM建立动态异常检测模型动态地追踪正常SWIM用户的浏览行为；然后通过学习和预测正常SWIM用户行为得出正常检测区间；最后选取访问包的大小和请求时间间隔为特征进行建模，并训练模型进行异常检测。实验结果表明，所提方法在攻击1和攻击2情况下检测率分别为99.95%和91.89%，与快速前向后向算法构建的HSMM相比，检测率提升了0.9%。测试结果表明所提方法可以有效地检测SWIM系统应用层DDoS攻击。     

基于隐马尔可夫模型的引文信息提取

提出一种基于符号特征提取的HMM结构学习方法，并利用修改的Viterbi算法进行引文信息提取。在实验结果的评价上，除了常用的准确率(Precision)、查全率(Recall)、综合评价指标(F)之外，引入了两个评价标准：宏平均和微平均来综合评价算法的性能。试验表明该方法用于引文信息提取的正确率较高。     

基于用户兴趣的个性化模型研究

大多数搜索引擎没有考虑到用户的个性和兴趣,大大降低了搜索的准确性。采用Web挖掘技术对存放在Web缓存中的历史页面进行挖掘,获取用户的兴趣信息,使用最优二叉树的形式来表示用户兴趣,利用获取的用户兴趣信息来构建个性化模型,并且利用智能Agent跟踪用户的兴趣变化,不断地对用户兴趣个性化模型进行更新。     

A novel approach to equipment health management based on auto-regressive hidden semi-Markov model （AR-HSMM）

As a new maintenance method, CBM （condition based maintenance） is becoming more and more important for the health management of complicated and costly equipment. A prerequisite to widespread deployment of CBM technology and prac- tice in industry is effective diagnostics and prognostics. Recently, a pattern recog- nition technique called HMM （hidden Markov model） was widely used in many fields. However, due to some unrealistic assumptions, diagnositic results from HMM were not so good, and it was difficult to use HMM directly for prognosis. By relaxing the unrealistic assumptions in HMM, this paper presents a novel approach to equip- ment health management based on auto-regressive hidden semi-Markov model （AR-HSMM）. Compared with HMM, AR-HSMM has three advantages： 1） It allows explicitly modeling the time duration of the hidden states and therefore is capable of prognosis. 2） It can relax observations＇ independence assumption by accom- modating a link between consecutive observations. 3） It does not follow the unre- alistic Markov chain＇s memoryless assumption and therefore provides more pow- erful modeling and analysis capability for real problems. To facilitate the computa- tion in the proposed AR-HSMM-based diagnostics and prognostics, new forward- backward variables are defined and a modified forward-backward algorithm is de- veloped. The evaluation of the proposed methodology was carried out through a real world application case study： health diagnosis and prognosis of hydraulic pumps in Caterpillar Inc. The testing results show that the proposed new approach based on AR-HSMM is effective and can provide useful support for the decision- making in equipment health management.     

主机型异常检测的隐半马尔可夫模型方法

提出基于HSMM模型的主机型入侵检测系统框架。以BSM审计数据作为数据源,提取正常主机行为的特权流系统调用序列,利用HSMM模型对正常主机行为进行建模,然后将当前主机行为与之比较,判定当前主机行为是否异常。选取特权流变化事件作为研究对象以缩短建模时间,同时滤去了过多的无用信息,一定程度上提高了检测效率。实验结果表明,提出的HSMM方法比HMM优越,同时该方法建模的系统不仅节省训练时间,而且在提高检测率的同时可以降低误报率。     

Hidden Markov model based filtering for singular semi-Markov jump systems

This article investigates the hidden Markov model based filter design problem for the singular semi-Markov jump systems (SSMJSs). The considered semi-Markov process is a generalization of Markov process, which can eliminate the restriction on the exponential distribution of sojourn time. Besides, the hidden Markov model based filter is introduced to tackle the asynchronous phenomenons occurred between the system modes and filter modes. To ensure the stochastic stability of the SSMJSs and derive solvable filter parameters, a filter design technic is constructed. First, the direct evolution of the states between two arbitrary close time instants is constructed from the filtering error system according to slow-fast decomposition, sufficient conditions are then proposed based on the consistent projector of the filtering error system and the constructed direct state evolution. Second, a new linear decoupling strategy is presented to deal with the coupled terms under the established stability conditions, which further derives the desired hidden Markov model based filter parameters. A numerical example is given to illustrate the effectiveness of the proposed method.     

Equipment PHM using non-stationary segmental hidden semi-Markov model

Health monitoring and prognostics of equipment is a basic requirement for condition-based maintenance (CBM) in many application domains where safety, reliability, and availability of the systems are considered mission critical. As a key complement to CBM, prognostics and health management (PHM) is an approach to system life-cycle support that seeks to reduce/eliminate inspections and time-based maintenance through accurate monitoring, incipient faults. Conducting successful prognosis, however, is more difficult than conducting fault diagnosis. A much broader range of asset health related data, especially those related to the failures, shall be collected. The asset health progression can then be possibly extracted from the congregated data, which has proved to be very challenging. This paper presents a non-stationary segmental hidden semi-Markov model (NSHSMM) based prognosis method to predict equipment health. Unlike previous HSMMs, the proposed NSHSMM no longer assumes that the state-dependent transition probabilities keep the same value all the time. That is, the probability of transiting to a less healthy state does not increase with the age. “Non-stationary” means the transition probabilities will change with time. In the proposed method, in order to characterize a deteriorating equipment, three kinds of aging factor that discount the probabilities of staying at current state while increasing the probabilities of transitions to less healthy states are introduced. The performances of these aging factors are compared by using historical data colleted from three hydraulic pumps. The hazard function (h.f.) has been introduced to analyze the distribution of lifetime with a combination of historical failure data and on-line condition monitoring data. Using h.f., PHM is based on a failure rate that is a function of both the equipment age and the equipment conditions. The state values of the equipment condition considered in PHM, however, are limited to those stochastically increasing over time and those having non-decreasing effect on the hazard rate. The estimated state duration probability distributions can be used to predict the remaining useful life of the systems. With the equipment PHM, the behavior of the equipment condition can be predicted.     

基于用户兴趣的跨网络用户身份识别算法

针对现有算法对用户兴趣在跨网络用户身份识别中作用的忽视以及时间复杂度高的问题,提出了基于用户兴趣的跨社交网络用户身份识别算法（UI-UI）。首先利用分块思想对用户节点进行初筛选,以提升算法效率、降低时间复杂度;其次,根据用户产生内容（UGC）和用户社交关系对用户兴趣进行建模,并计算兴趣相似度作为身份识别的依据;最后利用半监督学习的方法进行跨网络用户身份识别。通过在真实社交网络中进行实验,结果表明UI-UI算法能有效识别跨网络用户,且准确率和召回率稳定,运行时间显著减少。     

基于连续隐马尔可夫模型的人脸识别方法

提出了一种基于连续隐马尔可夫模型的人脸图像识别方法,主要内容包括以下方面:①由于奇异值向量具有稳定性.转置不变性等特点,对归一化的人脸图像,采用奇异值分解抽取人脸图像特征作为观察值序列;②在人脸识别中应用连续隐马尔可夫模型,采用双高斯概率密度函数训练,建立HMM模型,再利用建好的HMM模型进行识别.实验结果显示,所提出的方法减少了数据计算量,运行速度快,并提高了识别率,完全满足人脸识别系统实时性要求.     

基于点对分布的三维模型特征提取算法

基于形状分布算法提出了一种基于点对分布的三维模型特征提取算法。该算法根据两随机点间的连线与模型是否有交点,将连线分为U类和W类,并分别为两类连线构建形状分布直方图,从而解决了形状分布算法对于外形相近但细节不同的三维模型的检索问题,提高了检索精度。     

混合遗传算法和隐马尔可夫模型的Web信息抽取

传统Web信息抽取的隐马尔可夫模型对初值十分敏感和在实际训练中极易得到局部最优模型参数。提出了一种使用遗传算法优化HMM模型参数的Web信息抽取混合算法。该算法使用实数矩阵编码表示染色体,似然概率值为适应度取值,将GA与Baum-Welch算法相结合对HMM模型参数进行全局优化,并且调整GA-HMM的Baum-Welch算法参数实现Web信息抽取。实验结果表明,新的算法在精确度和召回率指标上比传统HMM具有更好的性能。     

Stylized facts of financial time series and hidden semi-Markov models

Hidden Markov models reproduce most of the stylized facts about daily series of returns. A notable exception is the inability of the models to reproduce one ubiquitous feature of such time series, namely the slow decay in the autocorrelation function of the squared returns. It is shown that this stylized fact can be described much better by means of hidden semi-Markov models. This is illustrated by examining the fit of two such models to 18 series of daily sector returns.     

基于用户兴趣特征提取的推荐算法研究*

传统的推荐算法一定程度上降低了网络消费者的搜索成本,但难以实时提供消费者满意的推荐服务,也忽略了用户偏好动态转移性。为了提高电子商务系统的推荐质量,从用户偏好的行为特征入手,建立了网络用户的兴趣特征提取模型,并设计了相应的推荐算法。通过对用户兴趣特征提取模型的检验和用户兴趣度矩阵的建立,依据与目标用户偏好相似的邻居用户对商品的兴趣程度预测用户对未浏览商品的兴趣度,并选择兴趣度值较高的N个商品推荐给用户。实验结果表明,在用户偏好动态转移的情况下,所设计的推荐算法的推荐精度和推荐效率明显提高,提高了网络用户的