云制造平台资源需求的高效匹配策略研究

为快速、精确地从云制造平台资源池中搜索到满足用户需求的资源,实现资源与需求的高效匹配,提出了一种资源需求高效匹配策略。首先,建立了资源与需求的形式化描述模型,在此基础上,采用改进的K-means聚类算法按基本信息进行聚类,对云服务池中的资源进行预处理,形成多个资源簇;其次,计算用户需求与各资源簇聚类中心基本信息的相似度,确立备选资源簇;最后,再分别从资源的状态信息、功能信息和服务信息3个方面对备选资源簇中的备选资源进行筛选匹配。实例分析和研究结果表明:与已有的匹配方法相比,该方法在保持较高匹配精度的同时具有更高的匹配效率。     

基于K-均值聚类与贝叶斯判别的我国煤矿顶板灾害事故安全评价

顶板事故是我国煤矿安全生产中的重大隐患,开展顶板灾害事故安全评价,识别顶板灾害危险性是煤矿安全生产亟待解决的问题。本文在文献调研的基础上,对影响顶板灾害的影响因素进行分析,从自然因素、技术因素和管理因素三个方面构建了顶板灾害安全评价指标体系,结合专家意见对各评价指标进行量化,采用K-均值聚类与贝叶斯判别的方法对我国顶板灾害安全水平进行评价,以我国煤矿实际样本数据为研究对象开展实证分析。研究结果表明:基于K-均值聚类与贝叶斯判别的煤矿顶板安全评价方法具有较高的准确度,能够用于定量化描述煤矿顶板灾害危险,具有一定的应用前景。     

改进模糊划分聚类的协同过滤推荐算法

针对传统协同过滤（CF）推荐算法存在评分矩阵稀疏、扩展性弱和推荐准确率低的缺陷，提出一种改进模糊划分聚类的协同过滤推荐算法（GIFP-CCF+）。在传统基于修正余弦相似度计算方法上，引入时间差因子、热门物品权重因子以及冷门物品权重因子以改善相似度计算结果；同时引入改进模糊划分的GIFP-FCM算法，将属性特征相似的项目聚成一类，构造索引矩阵，同索引间根据项目间的相似度寻找项目最近邻居构成推荐，从而提高协同过滤算法（CF）的精度。通过与Kmeans-CF、FCM-CF和GIFP-CCF算法进行仿真对比实验，证明了GIFP-CCF+算法在推荐结果和推荐精度上具有一定的优越性。     

铝电解关键指标预测方法的研究与应用

为维持铝电解生产的持续性，保证电解槽的物耗稳定和能耗稳定，通过对铝厂数据的挖掘与建模，提出一整套维持电解槽稳定的策略方法，并用于指导实际生产。首先为数据去噪，针对铝厂数据分布特征未知的特点，提出一种无参自适应的模糊聚类方法，通过迭代自适应得到类簇个数和簇中心；根据聚类结果，将铝厂数据按实际意义标签化，提出一种基于距离的连续属性朴素贝叶斯算法，对分类器使用增量思想，使算法动态分类准确率得到提高；应用单槽测试集数据，通过累积法完成当天各指标等级趋势的预测，确定各指标下变量相对于前一天的变化量，完成预测。实验发现，预测模型可完成铝电解关键指标的预测；提出的聚类、分类算法在UCI数据及铝厂数据上表现良好。     

基于NR-Transformer的集群作业运行时间预测

高性能集群的作业调度通常使用作业调度系统来实现，准确填写作业运行时间能在很大程度上提升作业调度效率。现有的研究通常使用机器学习的预测方式，在预测精度和实用性上还存在一定的提升空间。为了进一步提高集群作业运行时间预测的准确率，考虑先对集群作业日志进行聚类，将作业类别信息添加到作业特征中，再使用基于注意力机制的NR-Transformer网络对作业日志数据建模和预测。在数据处理上，根据与预测目标的相关性、特征的完整性和数据的有效性，从历史日志数据集中筛选出7维特征，并按作业运行时间的长度将其划分为多个作业集，再对各作业集分别进行训练和预测。实验结果表明，相比于传统机器学习和BP神经网络，时序神经网络结构有更好的预测性能，其中NR-Transformer在各作业集上都有较好的性能。     

Modified transform-based gamma correction for MRI tumor image denoising and segmentation by optimized histon-based elephant herding algorithm

Medical image processing is typically performed to diagnose a patient's brain tumor prior to surgery. In this study, a technique in denoising and segmentation was developed to improve medical image processing. The proposed approach employs multiple modules. In the first module, the noisy brain tumor image is transformed into multiple low- and high-pass tetrolet coefficients. In the second module, multiple low-pass tetrolet coefficients are applied through a modified transform-based gamma correction method. Generalized cross-validation is used on multiple high-pass tetrolet coefficients to obtain the best threshold value. In the third module, all enhanced coefficients are applied to the partial differential equation method. In the final module, the denoised image is applied to Atanassov's intuitionistic fuzzy set histon-based fuzzy clustering method with centroid optimization using an elephant herding method. Accordingly, the tumor part is segmented from the nontumor part in the magnetic resonance imaging brain images. The method was assessed in terms of peak signal-to-noise ratio, mean square error, specificity, sensitivity, and accuracy. The experimental results showed that the suggested method is superior to traditional methods.     

Machine learning–based piecewise affine model of wind turbines during maximum power point tracking

In this paper, a discrete‐time piecewise affine (PWA) model of a wind turbine during Maximum Power Point Tracking (MPPT) region is identified. A clustering‐based identification method is utilized to create PWA maps for nonlinear aerodynamic torque and thrust force functions. This method exploits the combined use of clustering, pattern recognition, and parameter identification techniques. The well‐known K‐means clustering method is employed along with a perceptron‐based multiclassifier for pattern recognition and the least squared technique for parameter estimation. The identified maps are approximated the nonlinear static functions of the dynamic model of the wind turbine. Characteristics of a 5‐MW wind turbine are considered and the resulting model, which consists of 25 subregions is compared with the nonlinear dynamic model. Two test cases are studied in order to validate the presented model. Simulation results demonstrate the effectiveness and accuracy of the PWA model such that the response of the identified PWA model is fitted well to the nonlinear one. The PWA model identified in this paper can be widely used for advanced control systems design and long‐term performance and security assessment of the power grid.     

基于SIR模型的隐蔽信道数据安全检测仿真

隐蔽信道数据分布散乱,对数据检测造成阻碍。针对传统的隐蔽信道数据检测方法存在检测速度慢、有效性差等问题,提出一种基于SIR模型的隐蔽信道数据安全检测方法。构建SIR隐蔽信道模型,使用在线检测模型进行隐蔽信道数据编码处理,使用密度聚类算法对隐蔽信道编码数据进行搜索聚类,划分密度区域,通过判断各密度区域数据有效性,完成隐蔽信道数据的密度聚类。利用决策树对聚类完成的数据进行特征属性提取,引入特征属性获取新的信息递增率,通过数据间差异性计算完成隐蔽信道数据安全检测。实验结果表明,所提方法能有效完成隐蔽信道数据检测,精准度、效率和稳定性均优于传统方法,且检测耗时少,具有显著优势。     

基于流数据聚类算法的电力大数据异常检测

针对电力大数据流的异常检测问题,该文将流数据聚类算法与电力大数据相结合,针对现有流数据聚类算法不易存储全部数据、断电数据易丢失等问题,以及流数据聚类算法对于离线阶段聚类算法实时应答的要求,从数据的完整性、安全性以及流数据聚类算法的低时间复杂度的角度出发,对CluStream流数据聚类算法进行改进,提出流式K-means聚类算法。对在线阶段,使用Redis集群进行流数据的缓冲,并设计节点时间衰减策略,增大心跳消息中有效消息所占比例;对离线阶段聚类算法进行优化,使用最佳距离法确定初始聚类中心,减少迭代次数;最后,使用所提出的流式K-means聚类算法进行用户用电异常行为检测,实验结果表明,该算法能够很好的发现用户用电异常行为。     

Ensemble cluster pruning via convex-concave programming

Ensemble learning is the process of aggregating the decisions of different learners/models. Fundamentally, the performance of the ensemble relies on the degree of accuracy in individual learner predictions and the degree of diversity among the learners. The trade-off between accuracy and diversity within the ensemble needs to be optimized to provide the best grouping of learners as it relates to their performance. In this optimization theory article, we propose a novel ensemble selection algorithm which, focusing specifically on clustering problems, selects the optimal subset of the ensemble that has both accurate and diverse models. Those ensemble selection algorithms work for a given number of the best learners within the subset prior to their selection. The cardinality of a subset of the ensemble changes the prediction accuracy. The proposed algorithm in this study determines both the number of best learners and also the best ones. We compared our prediction results to recent ensemble clustering selection algorithms by the number of cardinalities and best predictions, finding better and approximated results to the optimum solutions.