Attribute Reduction for Information Systems via Strength of Rules and Similarity Matrix

An information system is a type of knowledge representation,and attribute reduction is crucial in big data,machine learning,data mining,and intelligent systems.There are several ways for solving attribute reduction problems,but they all require a common categorization.The selection of features in most scientific studies is a challenge for the researcher.When working with huge datasets,selecting all available attributes is not an option because it frequently complicates the study and decreases performance.On the other side,neglecting some attributes might jeopardize data accuracy.In this case,rough set theory provides a useful approach for identifying superfluous attributes that may be ignored without sacrificing any significant information;nonetheless,investigating all available combinations of attributes will result in some problems.Furthermore,because attribute reduction is primarily a mathematical issue,technical progress in reduction is dependent on the advancement of mathematical models.Because the focus of this study is on the mathematical side of attribute reduction,we propose some methods to make a reduction for information systems according to classical rough set theory,the strength of rules and similarity matrix,we applied our proposed methods to several examples and calculate the reduction for each case.These methods expand the options of attribute reductions for researchers.

Clustering with Weighted Hyperlink and Sub Similarity Matrix

A web page clustering algorithm called PageCluster and the improved algorithm ImPageCluster solving overlapping are proposed. These methods not only take the web structure and page hyperlink into account, but also consider the importance of each page which is described as in-weight and out-weight. Compared with the traditional clustering methods, the experiments show that the runtimes of the proposed algorithms are less with the improved accuracies.

Similarity matrix-based K-means algorithm for text clustering

K-means algorithm is one of the most widely used algorithms in the clustering analysis. To deal with the problem caused by the random selection of initial center points in the traditional al- gorithm, this paper proposes an improved K-means algorithm based on the similarity matrix. The im- proved algorithm can effectively avoid the random selection of initial center points, therefore it can provide effective initial points for clustering process, and reduce the fluctuation of clustering results which are resulted from initial points selections, thus a better clustering quality can be obtained. The experimental results also show that the F-measure of the improved K-means algorithm has been greatly improved and the clustering results are more stable.

A random forest algorithm based on similarity measure and dynamic weighted voting

The random forest model is universal and easy to understand, which is often used for classification and prediction. However, it uses non-selective integration and the majority rule to judge the final result, thus the difference between the decision trees in the model is ignored and the prediction accuracy of the model is reduced. Taking into consideration these defects, an improved random forest model based on confusion matrix (CM-RF)is proposed. The decision tree cluster is selectively constructed by the similarity measure in the process of constructing the model, and the result is output by using the dynamic weighted voting fusion method in the final voting session. Experiments show that the proposed CM-RF can reduce the impact of low-performance decision trees on the output result, thus improving the accuracy and generalization ability of random forest model.

Audio Segmentation via the Similarity Measure of Audio Feature Vectors

A formula to compute the similarity between two audio feature vectors is proposed, which can map arbitrary pair of vectors with equivalent dimension to [0,1）. To fulfill the task of audio segmentation, a self-similarity matrix is computed to reveal the inner structure of an audio clip to be segmented. As the final result must be consistent with the subjective evaluation and be adaptive to some special applications, a set of weights is adopted, which can be modified through relevance feedback techniques. Experiments show that satisfactory result can be achieved via the algorithm proposed in this paper.

Measurement of Similarity for Spatial Directions Between Areal Objects

Similarity for spatial directions plays an important role in GIS. In this paper, the conventional approaches are analyzed. Based on raster data areal objects, the authors propose two new methods for measuring similarity among spatial directions. One is to measure the similarity among spatial directions based on the features of raster data and the changes of distances between spatial objects, the other is to measure the similarity among spatial directions according to the variation of each raster cell centroid angle. The two methods overcome the complexity of measuring similarity among spatial directions with direction matrix model and solve the limitation of small changes in direction. The two methods are simple and have broader applicability.

Gershgorin and Rayleigh Bounds on the Eigenvalues of the Finite-Element Global Matrices via Optimal Similarity Transformations

The large finite element global stiffness matrix is an algebraic, discreet, even-order, differential operator of zero row sums. Direct application of the, practically convenient, readily applied, Gershgorin’s eigenvalue bounding theorem to this matrix inherently fails to foresee its positive definiteness, predictably, and routinely failing to produce a nontrivial lower bound on the least eigenvalue of this, theoretically assured to be positive definite, matrix. Considered here are practical methods for producing an optimal similarity transformation for the finite-elements global stiffness matrix, following which non trivial, realistic, lower bounds on the least eigenvalue can be located, then further improved. The technique is restricted here to the common case of a global stiffness matrix having only non-positive off-diagonal entries. For such a matrix application of the Gershgorin bounding method may be carried out by a mere matrix vector multiplication.

A New Definition of Intuitionistic Fuzzy Similarity Degree

As far as the problem of intuitionistic fuzzy cluster analysis is concerned, this paper proposes a new formula of similarity degree with attribute weight of each index. We conduct a fuzzy cluster analysis based on the new intuitionistic fuzzy similarity matrix, which is constructed via this new weighted similarity degree method and can be transformed into a fuzzy similarity matrix. Moreover, an example is given to demonstrate the feasibility and validity of this method.

An Expected Patch Log Likelihood Denoising Method Based on Internal and External Image Similarity

Nonlocal property is an important feature of natural images,which means that the patch matrix formed by similar image patches is low-rank.Meanwhile,learning good image priors is of great importance for image denoising.In this paper,we combine the image self-similarity with EPLL(Expected patch log likelihood)method,and propose an EPLL denoising model based on internal and external image similarity to improve the preservation of image details.The experiment results show that the validity of our method is proved from two aspects of visual and numerical results.

基于余弦相似度列置换的Q矩阵修正方法

国内外研究者已开发出多种有效的Q矩阵修正方法,但当Q矩阵错误率较高时,仍存在修正效果不佳的问题.该文将基于塔克一致性系数和余弦相似度的列置换方法融入4种Q矩阵修正方法(GDI、Hull、MLR-B和stepwise)中,并借助Q矩阵向量和元素正确率等指标来评价新方法的修正效果.蒙特卡罗(Monte Carlo)模拟研究结果表明:在各种条件组合下,4种Q矩阵修正方法经过列置换后的修正效果得到明显提升,特别是当Q矩阵错误率较高时效果更加显著.

基于TOPSIS法和波士顿矩阵的医院神经专科能力评估研究

目的通过对以神经科学集群建设为特色的三级甲等医院的神经专科能力进行评估,为国内医院的特色专科建设提供参考。方法从服务能力、技术能力、质量安全和服务效率4个维度建立神经外科、神经内科专科能力评估指标体系,通过优劣解距离(technique for order preference by similarity to ideal solution,TOPSIS)法纵向比较首都医科大学附属北京天坛医院神经专科2019—2023年发展趋势,并采用波士顿矩阵深入分析神经专科各亚专业建设情况。结果2019—2023年首都医科大学附属北京天坛医院神经专科TOPSIS综合得分指数呈上升趋势。神经外科以胶质瘤诊治为主的肿瘤专业1,在技术能力和质量安全方面得分指数最高;神经内科以脑血管病为主的亚专业,在技术能力和服务效率方面得分指数较高,以上两个亚专业在波士顿矩阵中均处于优势巩固区。结论2019—2023年首都医科大学附属北京天坛医院神经专科诊治能力不断提升。神经外科亚专业中,专科能力排名最高的为肿瘤专业1,诊疗技术难度较高,同时医疗质量负性事件发生率低。神经内科各亚专业中,脑血管专业诊疗技术难度高且服务高效。

基于融合矩阵的文本相似度计算实现检索结果聚类

目的/意义弥补医学文本语义表示方面的不足,实现PubMed数据库检索结果聚类。方法/过程采用Jaccard系数和TF-IDF构建融合矩阵方法,建立短语间、文档间、短语与文档内容间的相似性关系融合矩阵,训练聚类算法,将PubMed数据库检索结果集合分组,随后生成类别标签,描述每一类簇文档的含义。结果/结论基于融合矩阵的聚类效果较好,提取出描述类别的高频词能很好地区分类别含义,对检索结果文本聚类任务有效。

基于Transformer紧凑编码的局部近重复视频检测算法

针对现有局部近重复视频检测算法特征存储消耗大、整体查询效率低、提取特征时并未考虑近重复帧之间细微的语义差异等问题,文中提出了一种基于Transformer紧凑编码的局部近重复视频检测算法。首先,提出了一个基于Transformer的特征编码器,其学习了大量近重复帧之间细微的语义差异,可以在编码帧特征时对各个区域特征图引入自注意力机制,在有效降低帧特征维度的同时也提高了编码后特征的表示性。该特征编码器通过孪生网络训练得到,该网络不需要负样本就可以有效学习近重复帧之间的相似语义信息,因此无需沉重和困难的难负样本标注工作,使得训练过程更加简易和高效。其次,提出了一个基于视频自相似度矩阵的关键帧提取方法,可以从视频中提取丰富但不冗余的关键帧,从而使关键帧特征序列能够更全面地描述原视频内容,提升算法的性能,同时也大幅减少了存储和计算冗余关键帧带来的开销。最后,基于关键帧的低维紧凑编码特征,采用基于图网络的时间对齐算法,实现局部近重复视频片段的检测和定位。该算法在公开的局部近重复视频检测数据集VCDB上取得了优于现有算法的实验性能。

基于改进松弛嵌入空间的多视图聚类

针对传统聚类方法缺乏统一特征表示,存在保守性的缺陷,提出一种基于改进松弛嵌入空间的多视图聚类方法。在统一的框架下联合学习一个综合的潜在嵌入表示矩阵、全局相似矩阵和一个精确指标矩阵。进一步放松全局相似矩阵的约束,并在此基础上提出一种改进的松弛多视图聚类嵌入空间,使得该方法具有更低的计算复杂度和更多的数据点对之间的相关性。实验结果表明,该方法能够获得鲁棒性更强、准确度更高的聚类结果。

关于矩阵多项式Jordan标准形刻画

本文利用矩阵运算、矩阵相似关系及矩阵的秩,深化了Jordan矩阵的性质,并在此基础上刻画了矩阵Jordan标准形中Jordan块的个数及阶数,最后讨论了矩阵多项式Jordan标准形,充实了高等代数中Jordan标准形的结果.

Study of Sustained Release Phenylpropanolamine Hydrochloride Hydrophilic Matrix Tablets Containing Hydroxypropylmethylcellulose K100M and Carbopol 971P

选用HPMC K100M和卡波普971P为骨架材料，粉末直接压片法制备骨架片，考察释放度，反相高效液相色谱法检测盐酸苯丙醇胺（PPA·HCl）的浓度。释药曲线均符合Higuchi方程（R＞0．98，P＜0．01）。运用正交设计法，以美国市场的PPA·HCl缓释片Acutrim^R为对照，相似因子f2值为指标，筛选获得了最优处方。其工艺重现性合格。研制片在0．1mol·L^-1 HCl,H2O(pH6.5)，磷酸盐缓冲液（PBS）pH5.0,6.8和7.4的介质中，以及在0.1mol·L^-1HCl中释放2h，转移至PBS6．8中释放10h，相对于对照品的f2值为63－74，表明在各介质中两制剂的释药曲线相似。释药影响因素的考察结果表明：在本实验考察的范围内，骨架片在水中的释药速率与HPMC K100M和卡波普971P的用量呈负相关。HPMC K100M和卡波普971P的比例（保持聚合物总用量相同），硬脂酸镁量和骨架片硬度对释药速率无显著性影响。

电力通信领域的颠覆性技术识别方法研究

通过分析颠覆性技术现状、未来发展趋势,建立一个合理、高效的技术识别方法,以有效地识别技术领域的颠覆技术,是当今技术发展中的一个重要问题。针对目前颠覆性技术识别方法具有专业领域的主观性及局限性,提出基于专利演化分析的识别方法。首先,计算专利数据的相似度矩阵、新颖性矩阵和演化矩阵,并从颠覆性技术激进的新颖性、相对快速的增长性、突出的影响性等特征出发,对专利的指标值进行对比分析。然后,以电力通信领域为例,构建技术专利数据集。最后,对所识别出的具有颠覆性特征的量子技术进行未来发展分析。该方法可以客观、有效地完成对电力通信领域的颠覆性技术识别。识别结果有助于完成电力通信产业转型升级,解决基础设施的信息安全问题。

基于改进FairMOT的多目标跟踪算法

针对复杂环境下多目标跟踪出现漏检与数据关联算法不友好导致目标之间频繁发生切换等问题,提出了基于FairMOT框架的多目标跟踪算法MFMOT。设计了轻量化多支路注意力模块,利用通道分组降低复杂度,从三个维度进行特征增强,使网络筛选提取到特征信息。重识别分支采用PolyLoss损失函数,增强同类目标之间的语义信息以区分同类不同目标对象。提出多特征融合相似度矩阵,通过融合多种特征相似度矩阵得到最优的相似度矩阵,降低目标之间身份切换次数。实验结果表明,在MOT17与MOT20数据集中HOTA分别为61.5%与56.1%,相比原有FairMOT模型分别提升2.2个百分点与2.3个百分点。与此同时,将多特征融合相似度矩阵应用至与FairMOT相同模式的多目标跟踪方法中,HOTA、MOTA与IDF1均得到提升。

基于标签挖掘的个性化推荐算法

基于标签的推荐算法中存在两个主要缺陷,缺乏用户对于标签偏好值的量化,以及不同标签在用户使用中所占权重。为此提出一种从标签角度出发的个性化推荐算法。分析用户历史行为中使用过的标签,根据用户历史行为建立用户的标签兴趣模型,利用标签兴趣模型计算用户对不同标签的偏好值;统计用户的历史评分记录,计算不同标签所占权重;将两者进行线性组合,得出用户对标签的兴趣度。利用余弦相似度,计算用户偏好相似度,将用户偏好相似度引入到矩阵分解模型中,进行项目评分预测和推荐。实验结果表明,在MovieLens数据集上,该算法相比于传统算法LFM和SVD++在RMSE上分别降低了5.00%和1.41%,在MAE上分别降低了5.07%和1.00%。

改进高斯云模型及其在装备保障体系能力评估中的应用

针对云模型在评估方法使用过程容易出现雾化特性差、评估结果不能直接使用的问题,提出一种改进的高斯云模型。该模型针对相同指标的专家评估结果,利用相似关系矩阵对专家进行聚类并赋权,求解每类专家的评估高斯云模型,并将其与权重结合形成该评估指标的评估云模型;综合所有指标的评估云模型形成综合评估云;将综合评估云与云标尺进行比对得出评估结果。运用该方法对装备保障体系能力进行评估,并与传统云模型所得评估结果进行对比,证明了改进高斯云模型的可行性和有效性,为其在评估方法中的运用提供参考。