基于极端梯度提升树算法的图像属性标注

提出基于极端梯度提升树(eXtreme gradient boosting,XGBoost)算法的图像属性标注模型,以改善标注性能：提取图像局部二值模式(local binary patterns,LBP)、灰度纹理空间包络特征(Gist)、尺度不变特征变换(scale invariant feature transform,SIFT)、视觉几何组(visual geometry group,VGG)等特征,以准确刻画图像视觉内容;基于图像特征,采用XGBoost算法集成弱分类器为强分类器,完成图像属性标注;深入挖掘图像属性蕴含的深层语义,构建全新的、层次化的属性表示体系,以贴近人类客观认知;设计迁移学习策略并合理组合分类模型,进一步改善标注性能。试验表明：Gist特征能真实刻画图像视觉内容;执行基础迁移学习后,标注精准度比迁移学习前最优指标提升8.69%;执行混合型迁移学习后,合理组合分类模型,标注精准度比基础迁移学习的最优指标提升17.55%。模型有效地改善图像属性标注精度。     

测试驱动开发探究

测试驱动开发是与传统程序开发方式相反的一种新型程序设计方法。它以测试编码来推动程序设计,大大缩减了设计和编码的冗余度。介绍了测试驱动开发基本思想、本质,并对测试驱动开发过程给出了清晰的流程,总结了测试驱动开发的优点。     

敏捷开发:极限编程在管理信息系统开发中的实践探讨

极限编程是敏捷的和基于实践的软件开发方法学。通过介绍极限编程的特点及其在一个管理信息系统项目实际开发中的成功实践,探讨研究了极限编程对于中小型需求易变的软件开发项目应用的优势和不足。     

A Haze Feature Extraction and Pollution Level Identification Pre-Warning Algorithm

The prediction of particles less than 2.5 micrometers in diameter (PM2.5) in fog and haze has been paid more and more attention, but the prediction accuracy of the results is not ideal. Haze prediction algorithms based on traditional numerical and statistical prediction have poor effects on nonlinear data prediction of haze. In order to improve the effects of prediction, this paper proposes a haze feature extraction and pollution level identification pre-warning algorithm based on feature selection and integrated learning. Minimum Redundancy Maximum Relevance method is used to extract low-level features of haze, and deep confidence network is utilized to extract high-level features. eXtreme Gradient Boosting algorithm is adopted to fuse low-level and high-level features, as well as predict haze. Establish PM2.5 concentration pollution grade classification index, and grade the forecast data. The expert experience knowledge is utilized to assist the optimization of the pre-warning results. The experiment results show the presented algorithm can get better prediction effects than the results of Support Vector Machine (SVM) and Back Propagation (BP) widely used at present, the accuracy has greatly improved compared with SVM and BP.     

基于DBN-XGBDT的入侵检测模型研究

在分布均匀的海量数据情况下,现有的入侵检测模型均具备良好的检测性能。但网络中产生的海量入侵数据的分布通常具有不均衡特点,而大多数检测模型针对罕见攻击类型的检测率低。针对上述问题,提出了一种深度信念网络（Deep Belief Networks,DBN）融合极限梯度提升（eXtreme Gradient Boosting,XGBoost）基于决策树算法（Decision Tree,DT）的入侵检测模型（DBN-XGBDT）。该模型将预处理后的数据集输入深度信念网络中,实现对入侵检测数据的降维处理,将得到的特征数据根据攻击类别任两类为一组,通过XGBoost算法逐一构建梯度提升树并细化为二分类;最后运用控制变量法和XGBoost内置的交叉验证进行调参,择优调整模型参数,对未知网络攻击实现有效检测。基于NSL-KDD数据集对DBN-XGBDT模型与XGBoost、DBN-BP、DBN-MSVM等优越模型进行了检测实验。实验结果表明,DBN-XGBDT模型较上述3个单一、混合分类模型的正确率分别提升2.07个百分点、1.14个百分点,对U2R的检测率提升至75.37%,平均误报率降至56.23%,为入侵检测处理不均衡数据且提高对罕见攻击的检测性能提供了新方法。     

基于经验模态分解自回归组合模型的网络舆情预测

随着大数据时代的到来,网络舆情数据呈现信息量大和领域覆盖广等特征。面对复杂的网络舆情数据时,传统单一模型预测能力有限,不能对舆情趋势进行有效预测。针对此问题,提出一种基于经验模态分解-自回归（EMD-AR）改进的组合模型——EMD-ARXG模型,应用于复杂网络舆情的预测。该模型利用经验模态分解算法对时间序列进行分解,然后通过自回归模型对分解后的时间序列进行各自趋势拟合,建立子模型。最后再对各个子模型进行重构,完成建模。另外,在利用自回归（AR）模型拟合过程中,为了减少拟合误差,采用极限梯度提升算法对残差进行学习,并使预测模型迭代更新,提高各个子模型预测精度。为验证EMD-ARXG模型的预测效果,该模型与小波神经网络模型和基于经验模态分解的神经网络模型进行实验对比。实验结果表明,在均方根误差（RMSE）、平均绝对百分误差（MAPE）和希尔不等系数（TIC）三项指标上,EMD-ARXG模型获得的结果均优于小波神经网络模型和基于经验模态分解的神经网络模型的结果。     

面向5G MEC基于行为的用户异常检测方案

5G边缘计算靠近用户侧提供服务,而边缘侧汇聚着用户的敏感信息,用户非法接入或合法用户自身的恶意行为威胁到整个边缘网络的安全。将机器学习算法应用于边缘计算架构,提出一种基于行为的用户异常检测方案。对用户行为进行建模,采用独热编码和互信息进行数据预处理和特征选择,并利用极限梯度提升算法训练一个多分类器分类识别进入园区的用户,根据识别结果与用户身份是否一致来判定用户是否异常。在此基础上,通过孤立森林算法对授权用户历史行为数据进行模型训练,从而检测可信任用户的行为是否异常,实现对小型固定园区内未授权用户的识别以及对授权用户异常行为的检测。实验结果表明,该方案可满足边缘计算场景的时间复杂度要求,并且能够有效区分不同用户,分类准确率达到0.953,而对异常行为样本的误报率仅为0.01。     

Collaborative simulation and scientific big data analysis: Illustration for sustainability in natural hazards management and chemical process engineering

Classical approaches for remote visualization and collaboration used in Computer-Aided Design and Engineering (CAD/E) applications are no longer appropriate due to the increasing amount of data generated, especially using standard networks. We introduce a lightweight and computing platform for scientific simulation, collaboration in engineering, 3D visualization and big data management. This ICT based platform provides scientists an “easy-to-integrate” generic tool, thus enabling worldwide collaboration and remote processing for any kind of data. The service-oriented architecture is based on the cloud computing paradigm and relies on standard internet technologies to be efficient on a large panel of networks and clients. In this paper, we discuss the need of innovations in (i) pre and post processing visualization services, (ii) 3D large scientific data set scalable compression and transmission methods, (iii) collaborative virtual environments, and (iv) collaboration in multi-domains of CAD/E. We propose our open platform for collaborative simulation and scientific big data analysis. This platform is now available as an open project with all core components licensed under LGPL V2.1. We provide two examples of usage of the platform in CAD/E for sustainability engineering from one academic application and one industrial case study. Firstly, we consider chemical process engineering showing the development of a domain specific service. With the rise of global warming issues and with growing importance granted to sustainable development, chemical process engineering has turned to think more and more environmentally. Indeed, the chemical engineer has now taken into account not only the engineering and economic criteria of the process, but also its environmental and social performances. Secondly, an example of natural hazards management illustrates the efficiency of our approach for remote collaboration that involves big data exchange and analysis between distant locations. Finally we underline the platform benefits and we open our platform through next activities in innovation techniques and inventive design.     

极限编程及其Android开发应用

介绍了极限编程的开发特点和方法，并给出了应用举例。     

基于极限编程的融合开发研究及其应用

文章提出了一种基于极限编程的多开发方法相融合的思路,并采用了自动化测试工具Junit.该测试工具在系统文件子模块的应用中提高了软件可维护性、可扩展性和开发速度.