基于随机森林模型的电信运营商外呼推荐系统

在电信运营商领域,外呼推荐是一种重要的推荐产品和服务的途径。实现了一种基于运营商大数据的自动外呼推荐系统,该系统能够挖掘用户的行为特征并且使用机器学习的方法预测用户对于被推荐产品的接受可能性。传统推荐系统使用的模型算法为矩阵分解、大规模稀疏特征分类、神经网络等。采用随机森林算法的主要原因是随机森林具有并行化程度高、训练速度快、生成的决策树可解释等诸多优点,适合于基于电信业数据的推荐系统。该外呼推荐系统基于Hadoop、Impala和Spark等大数据处理平台及工具,使用随机森林分类器作为核心算法,将用户最近的行为特征回归为接受外呼推荐产品的可能性。在线测试表明使用该系统与当前部署的人工随机外呼相比,能够提升约41%的用户接受率;同时,根据模型算法输出特征的重要性,进一步给出了两类用户的特征分析。     

基于无源化方法的移动机器人轨迹跟踪控制

该文将无源化方法应用到机器人轨迹跟踪控制中,根据机器人的动力学模型,采用基于模型及其转换矩阵的时变误差反馈的控制策略,将机器人的动力学过程转变为一个无损系统和关于一个二维向量的无源性映射。构造一个严格输入无源反馈映射,则此闭环系统为严格输出无源的,即此二维向量输出属于L2^n空间。由无源系统输出与轨迹跟踪误差间的转换矩阵形式可知跟踪误差也属于L2^n空间,即轨迹跟踪系统是稳定的。仿真结果表明,此方法控制效果良好,可以无差跟踪任意轨迹。     

基于特征组分层与半监督学习的鼠标轨迹识别

传统时间序列分类方法存在鼠标轨迹特征挖掘不充分、数据不平衡与标记样本量少等问题,造成识别效果较差。结合特征组分层和半监督学习,提出一种鼠标轨迹识别方法。通过不同视角构建有层次的鼠标轨迹特征组,并借鉴半监督学习的思想,利用多个随机森林模型对未标记样本进行伪标记,且将抽取标签预测一致且置信度较高的部分样本加入到训练集中。基于基础特征组和辅助特征组,在扩充后的训练集上训练随机森林模型,以实现鼠标轨迹的人机识别。实验结果表明,该方法可有效识别鼠标轨迹,且精确率、召回率与调和均值分别达到97.83%、94.72%和96.56%。     

Optimization-based iterative learning for precise quadrocopter trajectory tracking

Current control systems regulate the behavior of dynamic systems by reacting to noise and unexpected disturbances as they occur. To improve the performance of such control systems, experience from iterative executions can be used to anticipate recurring disturbances and proactively compensate for them. This paper presents an algorithm that exploits data from previous repetitions in order to learn to precisely follow a predefined trajectory. We adapt the feed-forward input signal to the system with the goal of achieving high tracking performance—even under the presence of model errors and other recurring disturbances. The approach is based on a dynamics model that captures the essential features of the system and that explicitly takes system input and state constraints into account. We combine traditional optimal filtering methods with state-of-the-art optimization techniques in order to obtain an effective and computationally efficient learning strategy that updates the feed-forward input signal according to a customizable learning objective. It is possible to define a termination condition that stops an execution early if the deviation from the nominal trajectory exceeds a given bound. This allows for a safe learning that gradually extends the time horizon of the trajectory. We developed a framework for generating arbitrary flight trajectories and for applying the algorithm to highly maneuverable autonomous quadrotor vehicles in the ETH Flying Machine Arena testbed. Experimental results are discussed for selected trajectories and different learning algorithm parameters.     

变参考轨迹下的鲁棒迭代学习模型预测控制

迭代学习模型预测控制是针对间歇过程的先进控制方法.它能通过迭代高精度跟踪给定参考轨迹,并保证时域上的闭环稳定性.然而,现有的迭代学习模型预测控制算法大多基于线性/线性化系统,且没有考虑参考轨迹变化的情况.本文基于线性参变系统提出一种能有效跟踪变参考轨迹的鲁棒迭代学习模型预测控制算法.首先,采用线性参变模型准确涵盖原始非线性系统的动态特性.然后,将鲁棒H_∞控制与传统迭代学习模型预测控制相结合,抑制变参考轨迹带来的跟踪误差波动,通过优化线性矩阵不等式约束下的目标函数求得控制输入.深入分析了鲁棒迭代学习模型预测控制的鲁棒稳定性和迭代收敛性.最后,通过对数值例子和连续搅拌反应釜系统的仿真验证了所提出算法的有效性.     

基于辅助变量的闭环系统子空间辨识

提出一种基于辅助变量的子空间辨识方法,适用于控制器信息未知以及参考输入已知的闭环系统参数辨识.通过将输入-输出数据块正交投影到辅助变量的行空间,直接得到扩展观测矩阵垂空间的估计.由此可从闭环系统中提取出对象模型信息,同时由SVD分解得到扩展观测矩阵与下三角Toeplitz矩阵的估计.给出了系统参数矩阵、噪声矩阵的计算方法.将所提出的子空间辨识方法应用于闭环动态的系统参数估计,其结果表明了该方法的有效性.     

基于ARF的Tor网站指纹识别技术

不法分子通过Tor等匿名通信系统构建暗网隐匿其不法行为,给网络监管带来了严峻挑战.网站指纹识别技术能根据加密流量来推测用户访问的站点,是一种有效的监管手段.已有的网站指纹识别技术采用的多为基于批处理的静态模型,无法有效解决概念漂移问题.针对Tor网站指纹,文章提出一种基于自适应随机森林(ARF)算法的动态网站指纹识别模...     

基于随机森林的高光谱遥感图像分类

为了充分利用高光谱图像的光谱信息和空间结构信息,提出了一种新的基于随机森林的高光谱遥感图像分类方法,首先,利用主成分分析降低数据的维数,并对主成分进行独立成分分析提取其光谱特征,同时消除像元的空间相关性,再采用形态学分析提取像元的空间结构特征,然后,根据像元的谱域和空域特征分别构造随机森林,并引入空间连续性对像元点的预测结果进行约束修正,最后由投票机制决定最后的分类结果。在AVIRIS和ROSIS高光谱图像上的实验结果表明,所提方法的分类性能要优于传统的高光谱图像分类方法,且分类精度高于基于单一特征的方法。     

基于遥感光谱和空间变量随机森林的黄河三角洲刺槐林健康等级分类

对刺槐林健康状况进行准确分类制图,是进行刺槐林健康状况评估与生态修复的前提。以高分辨率IKONOS影像、基于影像提取的不同窗口、不同灰度共生矩阵纹理信息以及反映局部空间自相关的Local Getis-Ord Gi(Getis统计量)为数据源,结合实测生态样方数据,利用多决策树的组合分类模型随机森林(RF)对刺槐林健康进行分级,对6种方法的分类精度进行了比较且对分类变量的重要性进行了排序。结果显示:19m×19m是最佳纹理计算窗口;灰度共生矩阵均值是最优纹理变量;基于波段4计算的Getis统计量对RF分类具有最重要的作用;较之利用全部光谱、纹理和Getis统计量的80个波段/变量,利用前向选择得到的前16个重要性变量进行RF分类,获得了最高的分类精度(总精度为93.14%,Kappa系数为0.894)。研究证实了从高分影像提取的空间特征信息有助于提高对具有规则分布格局的人工刺槐林健康等级的分类精度;前向选择方法可以利用较少的预测变量获得较高的分类精度。     

A linear differential operator approach to flatness based tracking for linear and non-linear systems

This contribution presents a flatness based solution to the tracking for linear systems in differential operator representation. Since the differential operator representation is a flat system representation, tracking controllers can easily be designed using dynamic output feedback. Then, the differential operator approach for flatness based tracking of linear systems is extended to non-linear systems. The design of the resulting linear time varying dynamic output feedback controller is based on a linearization about the trajectory, which directly yields the differential operator representation. Different from the non-linear flatness based controller design the new approach uses linear methods, both in stabilizing the tracking and in computing the output feedback controller. The proposed design procedure assures exact tracking in the steady state when no disturbances are present. A simple example demonstrates the design of a dynamic output feedback controller for the tracking of a non-linear system.     

The stochastic analysis of dynamic systems moving through random fields

In this paper we consider dynamic systems that move along specified trajectories across random fields, where the field acts as a driving force to the dynamic system. For a specific class of random fields we develop equations for the evolution of the covariance of the state of the dynamic system, and in the special case in which the trajectory is a straight line path followed by a 180° turn (i.e., an "over-and-back" trajectory) we develop a Markovian model that involves a change in the dimension of the state after the turn. For this case we also briefly discuss the estimation problem using recently developed results on "real-time smoothing."     

Development of expert control systems: A pattern classification and recognition approach

In this paper, a pattern classification and recognition approach to expert control systems is developed for use in the on-line analysis and design of dynamic systems. The approach used is based on the tuning of a three-term PID controller and, hence, it is not dependent on a specific form of the process model. A real-time experiment of implementing the developed controller using a microcomputer and associated hardware is presented. A sample set of production rules is discussed. The expert system reaches appropriate tuning parameters, using extracted features, such as oscillatory, underdamped, and exponentially monotonic properties.     

基于DNN的低资源语音识别特征提取技术

针对低资源训练数据条件下深层神经网络（Deep neural network,DNN）特征声学建模性能急剧下降的问题,提出两种适合于低资源语音识别的深层神经网络特征提取方法.首先基于隐含层共享训练的网络结构,借助资源较为丰富的语料实现对深层瓶颈神经网络的辅助训练,针对BN层位于共享层的特点,引入Dropout,Maxout,Rectified linear units等技术改善多流训练样本分布不规律导致的过拟合问题,同时缩小网络参数规模、降低训练耗时;其次为了改善深层神经网络特征提取方法,提出一种基于凸非负矩阵分解（Convex-non-negative matrix factorization,CNMF）算法的低维高层特征提取技术,通过对网络的权值矩阵分解得到基矩阵作为特征层的权值矩阵,然后从该层提取一种新的低维特征.基于Vystadial 2013的1小时低资源捷克语训练语料的实验表明,在26.7小时的英语语料辅助训练下,当使用Dropout和Rectified linear units时,识别率相对基线系统提升7.0%;当使用Dropout和Maxout时,识别率相对基线系统提升了12.6%,且网络参数数量相对其他系统降低了62.7%,训练时间降低了25%.而基于矩阵分解的低维特征在单语言训练和辅助训练的两种情况下都取得了优于瓶颈特征（Bottleneck features,BNF）的识别率,且在辅助训练的情况下优于深层神经网络隐马尔科夫识别系统,提升幅度从0.8%~3.4%不等.     

Information systems development as situated socio-technical change: a process approach

We outline a process approach for conducting qualitative research on how contemporary information systems (IS) development is enacted in practice. The approach involves a longitudinal case study to obtain rich data from which a detailed process narrative is produced. The narrative is informed by a conceptual model of IS development as situated socio-technical change, generating a theoretical explanation that highlights the temporal, emergent and contingent nature of IS development. This approach extends prior process studies of IS development by focusing on micro-level project activities, examining change as a continuous process and theorising IS change as the outcome of a dynamic trajectory of situated and socio-technical interactions. The methodological approach developed here can be used by other researchers to inform process studies of this complex organisational phenomenon.     

基于Spark平台的网络游戏用户流失预测方法

随着移动互联网的广泛普及,国内网络游戏市场日趋饱和,游戏公司获得新用户的成本不断增加,如何预防存量用户的流失已经成为市场营销的重心。提出了一种基于Spark平台的网络游戏用户流失预测方法,基于一个真实游戏日志数据对用户进行了流失预测。首先,从日志数据中抽取和计算了用户特征;随后,按权重选取了一组重要特征;最后,以特征为输入、流失与否为输出进行了二分类建模。综合比较了随机森林、支持向量机、多层感知机、梯度提升决策树和逻辑回归等6种常见分类算法。实验结果表明,随机森林算法表现最优,模型预测精度达到91%。     

Stabilization of Nonholonomic Robotic Systems Using Adaptation and Homogeneous Feedback

This paper considers the problem of stabilizing nonholonomic robotic systems in the presence of uncertainty regarding the system dynamic model. It is proposed that a simple and effective solution to this problem can be obtained by combining ideas from homogeneous system theory and adaptive control theory. Thus each of the proposed control systems consists of two subsystems: a (homogeneous) kinematic stabilization strategy which generates a desired velocity trajectory for the nonholonomic system, and an adaptive control scheme which ensures that this velocity trajectory is accurately tracked. This approach is shown to provide arbitrarily accurate stabilization to any desired configuration and can be implemented without knowledge of the system dynamic model. Moreover, it is demonstrated that exponential rates of convergence can be achieved with this methodology. The efficacy of the proposed stabilization strategies is illustrated through extensive computer simulations with nonholonomic robotic systems arising from explicit constraints on the system kinematics and from symmetries of the system dynamics.     

一种并行LSTM-FCN模型在船舶航迹预测中的应用

航迹预测是保障船舶航行安全、提高海洋交通管制效能、高效搜索海面目标的关键技术.为提高船舶航迹预测精确度,针对航迹特征多维度的特点,提出一种并行LSTM-FCN(parallel LSTM-FCN, PLSTM-FCN)模型.该模型有效结合LSTM模型对时间序列数据长期趋势预测的优势和全卷积网络(FCN)模型擅于提取时间序列数据细节变化规律的特点,通过并行结构设计保证相同训练效率下提取特征参数翻倍,实现较高精确度的高维航迹数据特征提取和趋势预测.基于动态时间规整算法和拉依达准则的船舶历史航迹数据预处理方法,可提高PLSTMFCN模型从不同类型船舶历史航迹中深度学习航行趋势和转弯细节的效率.基于船舶自动识别系统(AIS)数据的仿真实验结果表明, PLSTM-FCN模型对多维特征船舶航迹预测的精确度明显优于传统循环神经网络.     

基于长短期记忆网络的高速公路车辆变道轨迹预测模型

高速公路车辆车速、车距、行驶方向等因素都是动态变化的,受外界环境干扰,采集到的目标车辆状态特征数据可能存在噪声,导致车辆变道轨迹预测存在误差,为此提出基于长短期记忆网络的高速公路车辆变道轨迹预测模型,有效预测高速公路车辆变道轨迹,改善车辆行驶条件,保障其安全运行。通过激光雷达、GPS等装置采集目标车辆交通数据,将其合理组合成目标车辆状态观测特征向量,并构建相应的特征向量矩阵,将所构建目标车辆状态观测特征向量矩阵作为1层卷积神经网路输入,提取目标车辆状态观测特征向量潜在特征后,以1层卷积神经网络输出结果为双向长短期记忆网络有效输入,经过无数次模型训练后,输出目标车辆变道轨迹预测结果。实验结果表明:该模型可有效预测高速公路车辆变道轨迹,预测出的轨迹横纵坐标误差极低,能够得到较为理想的高速公路车辆变道轨迹预测结果。     

时空轨迹数据驱动的自动驾驶场景元建模方法

时空轨迹数据驱动的汽车自动驾驶场景建模,是当前汽车自动驾驶领域中驾驶场景建模、仿真所面临的关键问题,对于提高系统的安全性具有重要研究意义.近年来,随着时空轨迹数据建模及应用研究的快速发展,时空轨迹数据应用于特定领域建模的研究引起人们的广泛关注.但由于时空轨迹数据所反映现实世界的多元性和复杂性以及时空轨迹数据的海量、异构、动态等特点,基于时空轨迹数据驱动的安全攸关场景建模的研究仍面临着挑战,包括：统一的时空轨迹数据元模型、基于时空轨迹数据的元建模方法、基于数据分析技术的时空轨迹数据处理、数据质量评价等.针对汽车自动驾驶领域的场景建模需求,我们提出一种基于MOF元建模体系构建时空轨迹数据的元建模方法,根据时空轨迹数据的特征及自动驾驶的领域知识,构建了面向汽车自动驾驶的时空轨迹数据元模型;并基于此,提出基于时空轨迹数据元建模技术体系的自动驾驶安全场景建模方法,并使用场景建模语言ADSML实例化安全场景,构建安全场景库,旨在为此类系统的安全关键场景建模提供一种可行的方案.结合变道超车场景的案例,展示了时空轨迹数据驱动的自动驾驶安全场景元建模方法的可用性,为场景模型的构建、仿真、分析奠定了基础.     

嵌入式系统中视频运动对象分割

提出了一种基于嵌入式系统的视频运动对象分割算法。首先利用差图像法抽取出运动的像素点,然后通过统计像素点的状态变化频率来区分运动物体和动态背景,并配合一权值状态矩阵将全局光照突变和动态背景像素自适应融合到背景中,从而分割出运动对象并进行跟踪。实验结果表明,该算法在嵌入式系统中实时跟踪运动目标取得了很好的效果。