Diagnosis of Disc Space Variation Fault Degree of Transformer Winding Based on K-Nearest Neighbor Algorithm

Winding is one of themost important components in power transformers.Ensuring the health state of the winding is of great importance to the stable operation of the power system.To efficiently and accurately diagnose the disc space variation(DSV)fault degree of transformer winding,this paper presents a diagnostic method of winding fault based on the K-Nearest Neighbor(KNN)algorithmand the frequency response analysis(FRA)method.First,a laboratory winding model is used,and DSV faults with four different degrees are achieved by changing disc space of the discs in the winding.Then,a series of FRA tests are conducted to obtain the FRA results and set up the FRA dataset.Second,ten different numerical indices are utilized to obtain features of FRA curves of faulted winding.Third,the 10-fold cross-validation method is employed to determine the optimal k-value of KNN.In addition,to improve the accuracy of the KNN model,a comparative analysis is made between the accuracy of the KNN algorithm and k-value under four distance functions.After getting the most appropriate distance metric and kvalue,the fault classificationmodel based on theKNN and FRA is constructed and it is used to classify the degrees of DSV faults.The identification accuracy rate of the proposed model is up to 98.30%.Finally,the performance of the model is presented by comparing with the support vector machine(SVM),SVM optimized by the particle swarmoptimization(PSO-SVM)method,and randomforest(RF).The results show that the diagnosis accuracy of the proposed model is the highest and the model can be used to accurately diagnose the DSV fault degrees of the winding.

Characteristics,classification and KNN-based evaluation of paleokarst carbonate reservoirs:A case study of Feixianguan Formation in northeastern Sichuan Basin,China

The Feixianguan Formation reservoirs in northeastern Sichuan are mainly a suite of carbonate platform deposits.The reservoir types are diverse with high heterogeneity and complex genetic mechanisms.Pores,vugs and fractures of different genetic mechanisms and scales are often developed in association,and it is difficult to classify reservoir types merely based on static data such as outcrop observation,and cores and logging data.In the study,the reservoirs in the Feixianguan Formation are grouped into five types by combining dynamic and static data,that is,karst breccia-residual vuggy type,solution-enhanced vuggy type,fractured-vuggy type,fractured type and matrix type(non-reservoir).Based on conventional logging data,core data and formation microscanner image(FMI)data of the Qilibei block,northeastern Sichuan Basin,the reservoirs are classified in accordance with fracture-vug matching relationship.Based on the principle of cluster analysis,K-Nearest Neighbor(KNN)classification templates are established,and the applicability of the model is verified by using the reservoir data from wells uninvolved in modeling.Following the analysis of the results of reservoir type discrimination and the production of corresponding reservoir intervals,the contributions of various reservoir types to production are evaluated and the reliability of reservoir type classification is verified.The results show that the solution-enhanced vuggy type is of high-quality sweet spot reservoir in the study area with good physical property and high gas production,followed by the fractured-vuggy type,and the fractured and karst breccia-residual vuggy types are the least promising.

Computational Intelligence Prediction Model Integrating Empirical Mode Decomposition,Principal Component Analysis,and Weighted k-Nearest Neighbor

On the basis of machine leaning,suitable algorithms can make advanced time series analysis.This paper proposes a complex k-nearest neighbor(KNN)model for predicting financial time series.This model uses a complex feature extraction process integrating a forward rolling empirical mode decomposition(EMD)for financial time series signal analysis and principal component analysis(PCA)for the dimension reduction.The information-rich features are extracted then input to a weighted KNN classifier where the features are weighted with PCA loading.Finally,prediction is generated via regression on the selected nearest neighbors.The structure of the model as a whole is original.The test results on real historical data sets confirm the effectiveness of the models for predicting the Chinese stock index,an individual stock,and the EUR/USD exchange rate.

Fault Diagnosis in Robot Manipulators Using SVM and KNN

In this paper,Support Vector Machine(SVM)and K-Nearest Neighbor(KNN)based methods are to be applied on fault diagnosis in a robot manipulator.A comparative study between the two classifiers in terms of successfully detecting and isolating the seven classes of sensor faults is considered in this work.For both classifiers,the torque,the position and the speed of the manipulator have been employed as the input vector.However,it is to mention that a large database is needed and used for the training and testing phases.The SVM method used in this paper is based on the Gaussian kernel with the parametersγand the penalty margin parameter“C”,which were adjusted via the PSO algorithm to achieve a maximum accuracy diagnosis.Simulations were carried out on the model of a Selective Compliance Assembly Robot Arm(SCARA)robot manipulator,and the results showed that the Particle Swarm Optimization(PSO)increased the per-formance of the SVM algorithm with the 96.95%accuracy while the KNN algo-rithm achieved a correlation up to 94.62%.These results showed that the SVM algorithm with PSO was more precise than the KNN algorithm when was used in fault diagnosis on a robot manipulator.

简化的粒子群优化快速KNN分类算法

提出了一种有效的k近邻分类文本分类算法,即SPSOKNN算法,该算法利用粒子群优化方法的随机搜索能力在训练集中随机搜索,在搜索k近邻的过程中,粒子群跳跃式移动,掠过大量不可能成为k近邻的文档向量,并且去除了粒子群进化过程中粒子速度的影响,从而可以更快速地找到测试样本的k个近邻.通过验证算法的有效性表明,在查找k近邻相同时,SPOSKNN算法的分类精度高于基本KNN算法。

基于后验概率制导的B-KNN文本分类方法

针对K最近邻(KNN)方法分类准确率高但分类效率较低的特点,提出基于后验概率制导的贝叶斯K最近邻(B-KNN)方法。利用测试文本的后验概率信息对训练集多路静态搜索树进行剪枝,在被压缩的候选类型空间内查找样本的K个最近邻,从而在保证分类准确率的同时提高KNN方法的效率。实验结果表明,与KNN相比,B-KNN的性能有较大提升,更适用于具有较深层次类型空间的文本分类应用。

Outsmarting Android Malware with Cutting-Edge Feature Engineering and Machine Learning Techniques

The growing usage of Android smartphones has led to a significant rise in incidents of Android malware andprivacy breaches.This escalating security concern necessitates the development of advanced technologies capableof automatically detecting andmitigatingmalicious activities in Android applications(apps).Such technologies arecrucial for safeguarding user data and maintaining the integrity of mobile devices in an increasingly digital world.Current methods employed to detect sensitive data leaks in Android apps are hampered by two major limitationsthey require substantial computational resources and are prone to a high frequency of false positives.This meansthat while attempting to identify security breaches,these methods often consume considerable processing powerand mistakenly flag benign activities as malicious,leading to inefficiencies and reduced reliability in malwaredetection.The proposed approach includes a data preprocessing step that removes duplicate samples,managesunbalanced datasets,corrects inconsistencies,and imputes missing values to ensure data accuracy.The Minimaxmethod is then used to normalize numerical data,followed by feature vector extraction using the Gain ratio andChi-squared test to identify and extract the most significant characteristics using an appropriate prediction model.This study focuses on extracting a subset of attributes best suited for the task and recommending a predictivemodel based on domain expert opinion.The proposed method is evaluated using Drebin and TUANDROMDdatasets containing 15,036 and 4,464 benign and malicious samples,respectively.The empirical result shows thatthe RandomForest(RF)and Support VectorMachine(SVC)classifiers achieved impressive accuracy rates of 98.9%and 98.8%,respectively,in detecting unknown Androidmalware.A sensitivity analysis experiment was also carriedout on all three ML-based classifiers based on MAE,MSE,R2,and sensitivity parameters,resulting in a flawlessperformance for both datasets.This approach has substantial potential for real-world applications and can serve asa valuable tool for preventing the spread of Androidmalware and enhancing mobile device security.

Study on Chironomid Larvae Recognition Based on DWT and Improved KNN

A chironomid larvae images recognition method based on wavelet energy feature and improved KNN is developed. Wavelet decomposition and color information entropy are selected to construct vectors for KNN that is used to classify of the images. The distance function is modified according to the weight determined by the correlation degree between feature and class, which effectively improves classification accuracy. The result shows the mean accuracy of classification rate is up to 95.41% for freshwater plankton images, such as chironomid larvae, cyclops and harpacticoida.

水声目标识别中的K-D树KNN-SVM分类器研究

常规的KNN-SVM联合分类器中K-近邻算法没有充分挖掘训练样本的信息,使用遍历的方法来计算待识别样本与训练样本之间的距离,特别是在训练样本巨大时,存在大量的冗余计算。针对该问题,将训练样本训练成K-D树的结构,设计了K-D树KNN-SVM分类器,该分类器可以大大减少这些多余的计算,从而提高了搜索效率,有效缩短了搜索时间。进行了仿真和实验研究,分别设计了KNN、SVM、KNN-SVM分类器对两类水下目标进行了分类识别,并对相关参数的选取进行了优化。实验结果表明:在选定了最佳参数后的KNN-SVM联合分类器较其它两类分类器在识别率和识别效率方面都是最佳的;采用了K-D树结构的KNN-SVM联合分类器中KNN部分识别效率要比常规的高约7.5倍。

Using Deep Learning for Soybean Pest and Disease Classification in Farmland

To accurately identify soybean pests and diseases, in this paper, a kind of deep convolution network model was used to determine whether or not a soybean crop possessed pests and diseases. The proposed deep convolution network could learn the highdimensional feature representation of images by using their depth. An inception module was used to construct a neural network. In the inception module, multiscale convolution kernels were used to extract the distributed characteristics of soybean pests and diseases at different scales and to perform cascade fusion. The model then trained the SoftMax classifier in a uniformed framework. This realized the model of soybean pests and diseases so as to verify the effectiveness of this method. In this study, 800 images of soybean leaf images were taken as the experimental objects. Of these 800 images, 400 were selected for network training, and the remaining 400 images were used for the network test. Furthermore, the classical convolutional neural network was optimized. The accuracies before and after optimization were 96.25% and 95.81%, respectively, in terms of extracting image features. This type of research might be applied to achieve a degree of automation in agricultural field management.

A KNN-based two-step fuzzy clustering weighted algorithm for WLAN indoor positioning

Although k-nearest neighbors （KNN） is a popular fingerprint match algorithm for its simplicity and accuracy, because it is sensitive to the circumstances, a fuzzy c-means （FCM） clustering algorithm is applied to improve it. Thus, a KNN-based two-step FCM weighted （KTFW） algorithm for indoor positioning in wireless local area networks （WLAN） is presented in this paper. In KTFW algorithm, k reference points （RPs） chosen by KNN are clustered through FCM based on received signal strength （RSS） and location coordinates. The right clusters are chosen according to rules, so three sets of RPs are formed including the set of k RPs chosen by KNN and are given different weights. RPs supposed to have better contribution to positioning accuracy are given larger weights to improve the positioning accuracy. Simulation results indicate that KTFW generally outperforms KNN and its complexity is greatly reduced through providing initial clustering centers for FCM.

LF-CNN:Deep Learning-Guided Small Sample Target Detection for Remote Sensing Classification

Target detection of small samples with a complex background is always difficult in the classification of remote sensing images.We propose a new small sample target detection method combining local features and a convolutional neural network(LF-CNN)with the aim of detecting small numbers of unevenly distributed ground object targets in remote sensing images.The k-nearest neighbor method is used to construct the local neighborhood of each point and the local neighborhoods of the features are extracted one by one from the convolution layer.All the local features are aggregated by maximum pooling to obtain global feature representation.The classification probability of each category is then calculated and classified using the scaled expected linear units function and the full connection layer.The experimental results show that the proposed LF-CNN method has a high accuracy of target detection and classification for hyperspectral imager remote sensing data under the condition of small samples.Despite drawbacks in both time and complexity,the proposed LF-CNN method can more effectively integrate the local features of ground object samples and improve the accuracy of target identification and detection in small samples of remote sensing images than traditional target detection methods.

Using FCM to Select Samples in Semi-Supervised Classification

For a semi-supervised classification system, with the increase of the training samples number, the system needs to be continually updated. As the size of samples set is increasing, many unreliable samples will also be increased. In this paper, we use fuzzy c-means （FCM） clustering to take out some samples that are useless, and extract the intersection between the original training set and the cluster after using FCM clustering. The intersection between every class and cluster is reliable samples which we are looking for. The experiment result demonstrates that the superiority of the proposed algorithm is remarkable.

基于边际Fisher准则和迁移学习的小样本集分类器设计算法

如何利用大量已有的同构标记数据(源域)设计小样本训练数据(目标域)的分类器是一个具有很强应用意义的研究问题.由于不同域的数据特征分布有差异,直接使用源域数据对目标域样本进行分类的效果并不理想.针对上述问题,本文提出了一种基于迁移学习的分类器设计算法.首先,本文利用内积度量的边际Fisher准则对源域进行特征映射,提高源域中类内紧凑性和类间区分性.其次,为了筛选合理的训练样本对,本文提出一种去除边界奇异点的算法来选择源域密集区域样本点,与目标域中的标记样本点组成训练样本对.在核化空间上,本文学习了目标域特征到源域特征的非线性转换,将目标域映射到源域.最后,利用邻近算法(k-nearest neighbor,k NN)分类器对映射后的目标域样本进行分类.本文不仅改进了边际Fisher准则方法,并且将基于自适应样本对筛选的迁移学习应用到小样本数据的分类器设计中,提高域间适应性.在通用数据集上的实验结果表明,本文提出的方法能够有效提高小样本训练域的分类器性能.

云计算中保护数据隐私的快速多关键词语义排序搜索方案

可搜索加密技术主要解决在云服务器不完全可信的情况下,支持用户在密文上进行搜索.该文提出了一种快速的多关键词语义排序搜索方案.首先,该文首次将域加权评分的概念引入文档的评分当中,对标题、摘要等不同域中的关键词赋予不同的权重加以区分.其次,对检索关键词进行语义拓展,计算语义相似度,将语义相似度、域加权评分和相关度分数三者结合,构造了更加准确的文档索引.然后,针对现有的MRSE(Multi-keyword Ranked Search over Encrypted cloud data)方案效率不高的缺陷,将创建的文档向量分块,生成维数较小的标记向量.通过对文档标记向量和查询标记向量的匹配,有效地过滤了大量的无关文档,减少了计算文档相关度分数和排序的时间,提高了搜索的效率.最后,在加密文档向量时,将文档向量分段,每一段与对应维度的矩阵相乘,使得构建索引的时间减少,进一步提高了方案的效率.理论分析和实验结果表明:该方案实现了快速的多关键词语义模糊排序搜索,在保障数据隐私安全的同时,有效地提高了检索效率,减少了创建索引的时间,并返回更加满足用户需求的排序结果.

流形上的k最近邻分类方法

针对分类数据中存在噪声样本和维数问题,提出了流形上的k最近邻方法。首先,利用贝叶斯公式对经典k最近邻方法进行扩展,并采用核概率密度方法估计样本的局部联合概率密度;其次,建立噪声样本点对模型,并构建改进的边际本征图和相应的权值矩阵,通过定义目标函数寻找最优降维映射矩阵;最后,提出一个完整的流形上k最近邻算法。与6种经典方法在12个常用数据集上的实验比较表明,在大多数情况下所提方法的分类性能要优于其他方法。

数据规范化方法对K近邻分类器的影响

讨论最小-最大规范化、z-score规范化及小数定标规范化3种方法对K近邻分类器性能的影响,在12个标准UCI真实数据集和1个人工数据集上进行实验比较。实验结果表明,规范化方法在大部分数据集能上提高K近邻分类器的识别率。针对实验结果研究据规范化方法提升分类器性能的内在原因,给出根据数据属性的数值分布特点决定是否使用数据规范化方法的一般准则。

基于行走时脚摆角的步态识别方法

提出一种利用脚摆动特征进行步态识别的方法。对步态序列图像进行背景提取、图像差分、阈值分割、形态学后处理后,提取行走时的脚摆角作为特征参数,再分别采用BP神经网络、最近邻分类器和K近邻分类器法对这些特征数据进行识别分类与比较分析。实验结果表明,与同类方法相比,该方法可以更快速地进行步态识别,且识别性能较好。

基于粗糙集的微博用户性别识别

针对微博消息往往会不同程度表现出性别倾向性的特点,从消息内容挖掘的角度出发提出了一种基于粗糙集的微博用户性别识别算法。设计了一种基于容差粗集的微博消息表示模型(TRSRM),有效地刻画微博消息的性别特征。实验结果表明,在1000个真实微博用户的微博消息的测试集下,所提模型的准确率比特征项频数表示模型平均提高了7%,取得了更好的识别效果。

基于训练特征空间分布的雷达地面目标鉴别器设计

该文对雷达地面目标高分辨1维距离像目标识别中的库外目标鉴别问题,提出一种基于训练特征空间分布的雷达地面目标鉴别器。在训练阶段利用基于相关系数预处理的K-Means聚类方法对库内目标样本特征空间进行区域划分,并采用基于空间分布的支撑向量域描述方法确定样本特征空间的边界与支撑向量,利用样本特征空间边界与加权K近邻原则对目标类别进行判决。该方法解决了库内目标与库外目标的鉴别问题,提高了目标识别系统的总体性能。针对多种不同姿态下目标特征空间非均匀聚合的特点,对训练样本特征空间进行区域划分,减小模板匹配搜索运算规模,保证目标鉴别所需的实时性工作要求。最后通过仿真和实测数据验证了该方法具备优良的鉴别性能与良好的实时处理能力。