Prediction of elevator traffic flow based on SVM and phase space reconstruction

To make elevator group control system better follow the change of elevator traffic flow (ETF) in order to adjust the control strategy,the prediction method of support vector machine (SVM) in combination with phase space reconstruction has been proposed for ETF.Firstly,the phase space reconstruction for elevator traffic flow time series (ETFTS) is processed.Secondly,the small data set method is applied to calculate the largest Lyapunov exponent to judge the chaotic property of ETF.Then prediction model of ETFTS based on SVM is founded.Finally,the method is applied to predict the time series for the incoming and outgoing passenger flow respectively using ETF data collected in some building.Meanwhile,it is compared with RBF neural network model.Simulation results show that the trend of factual traffic flow is better followed by predictive traffic flow.SVM algorithm has much better prediction performance.The fitting and prediction of ETF with better effect are realized.

A Discrete-Time Traffic and Topology Adaptive Routing Algorithm for LEO Satellite Networks

“Minimizing path delay” is one of the challenges in low Earth orbit (LEO) satellite network routing algo-rithms. Many authors focus on propagation delays with the distance vector but ignore the status information and processing delays of inter-satellite links. For this purpose, a new discrete-time traffic and topology adap-tive routing (DT-TTAR) algorithm is proposed in this paper. This routing algorithm incorporates both inher-ent dynamics of network topology and variations of traffic load in inter-satellite links. The next hop decision is made by the adaptive link cost metric, depending on arrival rates, time slots and locations of source-destination pairs. Through comprehensive analysis, we derive computation formulas of the main per-formance indexes. Meanwhile, the performances are evaluated through a set of simulations, and compared with other static and adaptive routing mechanisms as a reference. The results show that the proposed DT-TTAR algorithm has better performance of end-to-end delay than other algorithms, especially in high traffic areas.

Comparison between Neural Network and Adaptive Neuro-Fuzzy Inference System for Forecasting Chaotic Traffic Volumes

This paper applies both the neural network and adaptive neuro-fuzzy inference system for forecasting short-term chaotic traffic volumes and compares the results. The architecture of the neural network consists of the input vector, one hidden layer and output layer. Bayesian regularization is employed to obtain the effective number of neurons in the hidden layer. The input variables and target of the adaptive neuro-fuzzy inference system are the same as those of the neural network. The data clustering technique is used to group data points so that the membership functions will be more tailored to the input data, which in turn greatly reduces the number of fuzzy rules. Numerical results indicate that these two models have almost the same accuracy, while the adaptive neuro-fuzzy inference system takes more time to train. It is also shown that although the effective number of neurons in the hidden layer is less than half the number of the input elements, the neural network can have satisfactory performance.

相空间重构与改进SMA优化SVR的网络流量预测

为提高网络流量预测精度,提出结合相空间重构与改进黏菌优化支持向量回归的预测模型。为解决黏菌算法收敛慢、易得局部最优的不足,引入3种形态对立学习对种群进行初始化,提高种群多样性;利用非线性反馈因子更新机制,均衡全局搜索与局部开发;设计柯西-高斯混合变异对最优解变异,扩展搜索空间,避免陷入局部最优。利用改进黏菌算法对支持向量回归优化调参,有效解决超参初值敏感缺陷,提高学习精度和收敛速度,以此构建网络流量预测模型。实验结果表明,改进模型预测误差更小,能够实现高精度和实时性预测要求。

Short-term traffic flow prediction with PSR-XGBoostconsidering chaotic characteristics

To improve the level of active traffic management,a short-term traffic flow prediction model is proposed by combining phase space reconstruction(PSR)and extreme gradient boosting(XGBoost)algorithms.Firstly,the traditional data preprocessing method is improved.The new method uses hierarchical clustering to determine the traffic flow state and fills in missing and abnormal data according to different traffic flow states.Secondly,one-dimensional data are mapped into a multidimensional data matrix through PSR,and the time series complex network is used to verify the data reconstruction effect.Finally,the multidimensional data matrix is inputted into the XGBoost model to predict future traffic flow parameters.The experimental results show that the mean square error,average absolute error,and average absolute percentage error of the prediction results of the PSR-XGBoost model are 5.399%,1.632%,and 6.278%,respectively,and the required running time is 17.35 s.Compared with mathematical-statistical models and other machine learning models,the PSR-XGBoost model has clear advantages in multiple predictive indicators,proving its feasibility and superiority in short-term traffic flow prediction.

Multi-Gradient Routing Protocol for Wireless Sensor Networks

Sensor networks tend to support different traffic patterns since more and more emerging applications have diverse needs. We present MGRP, a Multi-Gradient Routing Protocol for wireless sensor networks, which is fully distributed and efficiently supports endto-end, one-to-many and many-to-one traffic patterns by effectively construct and maintain a gradient vector for each node. We further combine neighbor link estimation with routing information to reduce packet exchange on network dynamics and node failures. We have implemented MGRP on Tiny OS and evaluated its performance on real-world testbeds. The result shows MGRP achieves lower end-to-end packet delay in different traffic patterns compared to the state of the art routing protocols while still remains high packet delivery ratio.

Load Control for Overloaded MPLS/DiffServ Networks during SLA Negotiation

In end-to-end QoS provisioning some bandwidth portions on the link may be reserved for certain traffic classes (and for particular set of users) so the congestion problem of concurrent flows (traversing the network simultaneously) can appear. It means that in overloaded and poorly connected MPLS/DS networks the CR (Constraint-based Routing) becomes insufficient technique. If traffic engineering is supported with ap-propriate traffic load control the congestion possibility can be predicted before the utilization of guaranteed service. In that sense the initial (proactive) routing can be pre-computed much earlier, possible during SLA (Service Level Agreement) negotiation. In the paper a load simulation technique for load balancing control purpose is proposed. It could be a very good solution for congestion avoidance and for better load-balancing purpose where links are running close to capacity. To be acceptable for real application such complicated load control technique needs very effective algorithm. Proposed algorithm was tested on the network with maximum M core routers on the path and detail results are given for N=3 service classes. Further improve-ment through heuristic approach is made and results are discussed. Some heuristic options show significant complexity savings that is appropriate for load control in huge networks.

基于MeAEG-Net的异常流量检测方法研究

异常流量检测现有方法大都是基于有监督的学习,在现实生活中获取并标记异常流量数据样本是极为困难的,存在诸多限制.此外,由于网络异常数据的多样性和复杂性,各种检测方法的自适应性较差,对新出现的异常流量难以判断.针对上述问题,本文设计了一个基于生成对抗网络和记忆增强模块的半监督异常流量检测框架MeAEG-Net(Memory Augment Based on Generative Adversarial Network),通过只训练正常流量样本数据,比较生成器模块输入流量底层特征的重构误差来达到检测异常的目的 .在模型中使用生成对抗网络来更好地训练生成器,生成器采用自编码器加解码器的结构来解决自编码器易受噪声影响的问题,并在自编码器子网络中添加记忆增强模块来削弱生成器模块的泛化能力,增大异常流量的重构误差.实验证明,本文提出的方法能在只学习正常流量数据样本的前提下达到很好的异常流量检测效果.

基于最优系统弹性的城市道路快速化改造施工时序研究

为科学合理地制订城市道路快速化改造施工时序方案,减小施工改造过程中对周边路网交通出行所造成的影响,研究了基于最优系统弹性的城市道路快速化改造施工时序问题。以路网吸引力作为路网性能的评价指标,构建了路网弹性评价模型,并构建了基于系统弹性最优的城市道路快速化改造施工时序方案决策双层规划模型。上层模型以路网系统弹性最大为优化目标,用于制订城市道路快速化改造施工时序方案,下层模型为固定需求下的系统均衡分配模型,用于模拟给定快速化改造施工时序方案下的路网交通流演变,并将路段交通信息反馈给上层模型。设计了基于Frank_Wolfe的人工蚁群算法对模型进行求解,利用Nguyen&Dupuis网络对模型和算法进行了验证,并对出行时间忍耐系数进行了敏感性分析。结果表明:基于系统弹性的双层规划模型可以有效调整道路快速化改造施工时序方案,使得道路施工改造对路网性能所造成的破坏达到最小,相对于传统的施工时序决策方法,基于系统弹性最优的施工时序决策优化模型能够提高路网的系统弹性,减少19.7%的路网性能损失;随着出行时间忍耐系数的逐渐增大,道路快速化改造过程中路网性能下降逐渐变小,路网系统弹性逐渐变大。

利用图像超分辨率提升交通标志分类精度研究

为了解决利用卷积神经网络进行交通标志分类时精度低的问题,通过将图像超分辨率网络与分类网络相连接,提出一种超分级联网络结构。首先使用改进的双重注意力机制超分辨率网络作为级联网络的子网络;然后训练图像分类网络,用于对超分辨率处理后的图像进行分类;最后利用分类准确率衡量超分辨率重构对图像分类任务的有效性。模拟和真实交通标志数据集验证结果表明,经过超分辨率处理的图像在分类模型中均取得了更高的分类准确率,证明了超分辨率技术对于交通标志图像分类准确率的提高具有促进作用。

一种基于P2P流量规划的网络资源可重构分配方法

现有IP网络的无连接性和网络资源的松散管理结构使得网络流量存在全局范围内的分布不均情况,造成网络资源利用率低并且对运营商的一些突发的QoS需求很难做到快速响应.具备网络资源集中式管理特征的可重构网络虽然弥补了这一不足,但对特定业务的流量优化能力以及按需合理重构网络资源的能力不足,使得可重构网络资源分配及QoS保障技术仍有待进一步提高,尤其是占据网络流量主体的P2P业务流量和突发性高优先级QoS保障.当前的P2P优化以及流量工程技术不但自身存在不完善,并且也不能高效地直接适用于可重构网络环境中.因此,文中站在网络资源全局统筹的角度,提出了一种应用于可重构网络环境中的、基于P2P流量规划的网络资源动态可重构技术,使得可重构网络的资源管理机制有能力对全网P2P流量进行优化调度,尤其是能够在网络资源不足的情况下,通过调整P2P流量,释放指定网络资源以响应突发高优先级需求.通过仿真验证可见,以P2P流量规划的手段对全网资源进行全局规划、动态重构,形成弹性的可重构网络环境中,网络流量负担、P2P节点出、入度等多方面均有较大幅度提升,从而赋予了运营商一种对各种优先级业务的规划手段.

基于小波神经网络的流量混沌时间序列预测

在Takens提出的相空间重构模型基础上,应用小波变换对其进行改进,充分考虑噪声对重构结果的影响。将小波神经网络混沌时间序列预测方法引入网络流量预测中,介绍小波神经网络的基本构造和学习方法。实验表明,与RBF神经预测方法相比,小波神经网络预测方法的逼近效果更好、误差更小。

基于混沌和RBF神经网络的短时交通流量预测

针对传统的应用数学模型方法在短时交通流预测精度和实时性方面存在的问题,论文从非线性时间序列的角度对短时交通流量预测进行探讨,提出采用基于混沌理论的RBF神经网络预测方法。首先在采用小数据量的Lyapunav指数计算方法判定交通流存在混沌的前提下,对交通流量数据进行相空间重构。构建了RBF神经网络,并对模拟产生的Lorenz和Rossler混沌时间序列数据以及实际采集的高速公路交通流量数据进行了仿真研究。结果表明,该方法对模拟产生的混沌时间序列具有很好的预测效果,在交通流量的短时预测上也具有较高的预测精度。

基于混沌特性的网络流量预测

高速网络中存在着以自相似为特征的多种业务流量,这种自相似特征和混沌现象的吸引子有着紧密的联系。本文基于混沌时间序列重构相空间理论,根据最大Lyapunov指数,分别采用W o lf原始算法和改进算法,对高速网络中自相似信源的速率进行了预测,并给出了最大可预报时间。仿真结果表明,W o lf改进算法预测精度及可靠性更高。

基于KPCA优化ESN的网络流量预测方法

为了提高网络流量的预测精确度,提出一种核主成分分析(KPCA)优化回声状态网络(ESN)的网络流量预测方法。首先利用相空间重构对网络流量序列进行处理,提高序列的可预测性,然后对网络流量序列进行核主成分分析,提取序列中的有效信息,通过实验方法确定回声状态网络的储备池参数,最后利用回声状态网络对网络流量进行预测。与标准回声状态网络、差分自回归滑动平均模型(ARIMA)、以及最小二乘支持向量机(LSSVM)预测模型进行了仿真对比,结果表明提出的方法具有更高的预测精确度以及更小的预测误差,同时一定程度上减少了预测时间。

基于小波和回声状态网络的交通流多步预测模型

针对交通流的含噪混沌特征,提出了一种基于小波回声状态网络的交通流多步预测模型。该模型利用小波多尺度分解方法,屏蔽了噪声成分对交通流动力学特性的干扰,同时提取了占有交通流绝大部分能量的混沌低频成分。在采用多路分量并行预测的方式下,充分发挥了回声状态网络对混沌低频分量的强大多步预测能力,从而保障了交通流多步预测的精度。对北京市西直门桥的实测交通流的预测结果表明:该模型的多步预测精度比传统的回声状态网络模型有了较大幅度的提升,在保证预测精度的前提下,最大可预测的步长也相应的增加。

基于小波和多维重构的BP神经网络交通流短时预测

针对采集过程中噪声影响以及交通流时间序列的强相关性,提出一种基于小波和多维重构的BP神经网络交通流短时预测方法。运用启发式小波降噪法对原始交通流数据进行降噪处理,剔除数据中的噪声;基于C-C法将交通流数据进行多维度相空间重构,充分挖掘交通流的多维变化特性;构建多维度的BP神经网络进行交通流短时预测研究。运用2 400组数据进行实验,并与传统的BP神经网络、Elman神经网络以及SVM进行对比。实验结果表明,该方法具有较高的预测精度,绝对误差降低约2.408 0,均方误差降低约26.597 2。

基于粗集理论和支持向量机的道路网短时交通流量预测

为了解决城市交通流量预测问题,研究道路网中多断面同时作用的情况,先将各断面交通流时间序列在相空间中重构,以充分提取交通流中的相关信息,然后应用粗集理论的强定性分析能力对输入信息进行约简,消除了样本中的噪声和冗余。在此基础上,再利用支持向量机对约简信息进行预测。为了获得最优的预测精度,该方法还利用遗传算法对预测进行了优化。实例研究表明,该方法的预测效果令人满意,在交通控制领域具有较大的应用潜力。

基于量子神经网络的道路交通事故预测

道路交通事故预测是道路交通安全研究的一项重要内容.针对BP神经网络在道路交通事故预测中精度不足及收敛速度慢的问题,引入量子神经网络并构建道路交通事故预测模型.模型通过对道路交通事故时间序列进行相空间重构,有效扩充训练样本数量;且隐含层神经元采用态叠加的激励函数,对道路交通事故数据的特征空间进行多层梯级划分,以快速匹配输入数据与特征空间的对应关系,提高模型的收敛速度;在训练过程中动态调整量子间隔,以响应事故数据的强随机性.实验结果表明,该预测模型能够较好地适应道路交通事故数据的特性,且预测精度和收敛速度较改进BP神经网络有显著提高.

基于PSR-LSSVM的网络流量预测

为了提高网络流量预测精度,利用相空间重构和预测模型参数间的相互联系,提出一种遗传优化最小二乘支持向量机的网络流量预测方法。首先将相空间重构和最小二乘支持向量机参数作为遗传算法的个体,将模型预测精度作为个体适应度函数,然后通过遗传操作获得模型全局最优参数,最后通过网络流量仿真实验进行性能测试。结果表明,相对于传统预测方法,遗传优化最小二乘支持向量机提高了网络流量的预测精度,为网络流量预测提供了一种新的研究思路。