GBDT与LR融合模型在加密流量识别中的应用

随着网络应用服务类型的多样化以及网络流量加密技术的不断发展,加密流量识别已经成为网络安全领域的一个重大挑战。传统的流量识别技术如深度包检测无法有效地识别加密流量,而基于机器学习理论的加密流量识别技术则表现出很好的效果。因此,本文提出一种融合梯度提升决策树算法(GBDT)与逻辑回归(LR)算法的加密流量分类模型,使用贝叶斯优化(BO)算法进行超参数调整,利用与时间相关的流特征对普通加密流量与VPN加密流量进行识别,实现了整体高于90%的流量识别准确度,与其他常用分类模型相比拥有更好的识别效果。     

Offline and online fault detection and diagnosis of induction motors using a hybrid soft computing model

In this paper, a hybrid soft computing model comprising the Fuzzy Min-Max (FMM) neural network and the Classification and Regression Tree (CART) for motor fault detection and diagnosis is described. Specifically, the hybrid model, known as FMM-CART, is used to detect and classify fault conditions of induction motors in both offline and online environments. A series of experiments is conducted, whereby the Motor Current Signature Analysis (MCSA) method is applied to form a database containing stator current signatures under different motor conditions. The signal harmonics from the power spectral density (PSD) are extracted, and used as the discriminative input features for fault classification with FMM-CART. Three main induction motor conditions, viz. broken rotor bars, stator winding faults, and unbalanced supply, are used to evaluate the effectiveness of FMM-CART. The results indicate that FMM-CART is able to detect motor faults in the early stage, in order to avoid further damage to the induction motor as well as the overall machine or system that uses the motor in its operations.     

A threshold insensitive method for locating the forest canopy top with waveform lidar

Lidars have the unique ability to make direct, physical measurements of forest height and vertical structure in much denser canopies than is possible with passive optical or short wavelength radars. However the literature reports a consistent underestimate of tree height when using physically based methods, necessitating empirical corrections. This bias is a result of overestimating the range to the canopy top due to background noise and failing to correctly identify the ground.This paper introduces a method, referred to as “noise tracking”, to avoid biases when determining the range to the canopy top. Simulated waveforms, created with Monte-Carlo ray tracing over geometrically explicit forest models, are used to test noise tracking against simple thresholding over a range of forest and system characteristics. It was found that noise tracking almost completely removed the bias in all situations except for very high noise levels and very low (< 10%) canopy covers. In all cases noise tracking gave lower errors than simple thresholding and had a lower sensitivity to the initial noise threshold.Finite laser pulses spread out the measured signal, potentially overriding the benefit of noise tracking. In the past laser pulse length has been corrected by adding half that length to the signal start range. This investigation suggests that this is not always appropriate for simple thresholding and that the results for noise tracking were more directly related to pulse length than for simple thresholding. That this effect has not been commented on before may be due to the possible confounding impacts of instrument and survey characteristics inherent in field data. This method should help improve the accuracy of waveform lidar measurements of forests, whether using airborne or spaceborne instruments.     

基于树核函数的英文句子情感分类研究

最近几年,由于在线客户评论信息飞快地增长。如何把这些信息分类为正向和负向情感是一个迫切需要解决的问题。提出了一种细粒度级别(句子级别)的情感分类方法,该方法在SVM分类器中使用了树核和复合核函数来进行句子级别情感的分类。实验结果表明在句子级别的情感分类中树核和复合核的方法比线性核具有更佳的性能。     

复小波域层内层间相关性图像去噪方法

提出了双树复小波变换域尺度内和尺度间复系数相关性图像去噪新方法。该方法利用双树复小波变换的多方向性和平移不变性对图像进行多尺度分解,采用邻域复系数微分窗对其高频方向子图进行尺度内复系数相关性建模,并按最小错误率贝叶斯决策规则进行分类和状态标识;再把复系数尺度内状态标识与复小波域隐马尔可夫树相结合,从而实现降噪功能。实验结果表明,该方法在峰值信噪比指标上优于传统的滤波方法,能有效地抑制噪声的同时,对图像边缘具有较好的保护能力。     

基于prüfer数的遗传算法求解度约束最小树问题

度约束最小树问题属于NP-完全问题,是一类比较难解的问题,但在现实中具有非常重要的应用价值。探讨了如何将基于prüfer数的遗传算法应用于该问题,并给出了相应的算法。采用C语言和MATLAB的混合编程实现该算法,数值分析的结果显示了遗传算法求解该问题的有效性及其应用价值。     

A Class of Algorithms for Collision Resolution with Multiplicity Estimation

The wireless connectivity in pervasive computing has ephemeral character and can be used for creating ad hoc networks, sensor networks, connection with RFID (Radio Frequency Identification) tags etc. The communication tasks in such wireless networks often involve an inquiry over a shared channel, which can be invoked for: discovery of neighboring devices in ad hoc networks, counting the number of RFID tags that have a certain property, estimating the mean value contained in a group of sensors etc. Such an inquiry solicits replies from possibly large number of terminals n. This necessitates the usage of algorithms for resolving batch collisions (conflicts) with unknown conflict multiplicity n. In this paper we present a novel approach to the problem of collision resolution for batch conflicts. We show how the conventional tree algorithms for collision resolution can be used to obtain progressively accurate estimation of the multiplicity. We use the estimation to propose a more efficient binary tree algorithm, termed Estimating Binary Tree (EBT) algorithm. The EBT algorithm is suited for implementation when the conflicting nodes are passive, such as e.g. RFID tags. We extend the approach to design the Interval Estimation Conflict Resolution (IECR) algorithm. For n→∞ we prove that the efficiency achieved by IECR for batch arrivals is identical with the efficiency that Gallager’s FCFS algorithm achieves for Poisson packet arrivals. For finite n, the simulation results show that IECR is, to the best of our knowledge, the most efficient batch resolution algorithm reported to date.     

Spanning tree-based genetic algorithm for bicriteria transportation problem

In this paper, we present a new approach which is spanning tree-based genetic algorithm for bicriteria transportation problem. The transportation problem have the special data structure in solution characterized as a spanning tree. In encoding transportation problem, we introduce one of node encoding which is adopted as it is capable of equally and uniquely representing all possible basic solutions. The crossover and mutation was designed based on this encoding. And we designed the criterion that chromosome always feasibility converted to a transportation tree. In the evolutionary process, the mixed strategy and roulette wheel selection is used. Numerical experiments will be shown the effectiveness and efficiency of the proposed algorithm.     

高光谱影像的BDT-SVM地物分类算法与应用

面对海量数据的特征空间高维性及训练样本的有限性,高光谱遥感影像若采用常规统计模式的分类方法难以获得较好的分类结果。因此探讨支持向量机(SVM)分类器的基本原理,针对EO-1Hyperion高光谱影像的分类特点及现有多类SVM算法所存在的训练时间长及分类精度低等问题,引入二叉决策树SVM(BDT-SVM)分类算法,并提出一种新的类间分离度定义方法及相应的客观确定二叉树结构的策略,由此生成改进的BDT-SVM算法。实验结果表明:与其他多类分类方法相比,基于改进的BDT-SVM算法的高光谱影像地物分类效果更好,总体精度达到90.96%,Kappa系数为0.89,该算法还解决了经典SVM多类分类可能存在的不可分区域问题。