基于时序特征和集成算法的用户购买预测

大数据时代,电商平台积累了大量用户在平台上的行为数据,比如浏览、点击、下单和加入购物车等等.如何使用机器学习算法去探索大数据背后的用户消费喜好和习惯成为了一个新的研究热点.本文主要在特征工程和模型搭建两个方面对用户购买预测的效果做出提高.通过深入理解电商业务知识,利用统计学知识,分别从用户、商品和评论数据等多个方面的数据构建了115个特征;在模型搭建方面,主要设计了一个两层融合模型,第一层采用了XGBoost、CatBoost和逻辑回归作为基分类器,从不同的角度考虑用户购买预测,第二层采用加权平均的方法对基类模型的预测结果进行融合,其权重由线性分类器学习生成.实验结果表明该融合模型的F1评分要高于个体分类器,并且多次实验证明,融合模型的稳定性也要比个体分类器好.     

基于多次散射目标模型的雷达成像研究

合成孔径雷达（Synthetic Aperture Radar, SAR）作为一种主动微波传感器,被广泛应用于遥感对地观测。然而,传统上的雷达散射模型是基于几何光学近似的,忽略了地物目标之间的相互作用,即多次散射效应。为了充分挖掘和利用毫米波雷达影像中目标电磁散射特性信息,亟需对典型目标的多次散射效应进行建模仿真和验证。基于矩量法（Method of Moments, MoM）分析了球体、二面角等目标表面等效电流分布与分段成像结果的对应关系,并利用后向投影（Back-Projection, BP）算法进行电磁仿真成像,总结了单/双基地雷达模式对成像中散射机制的影响。结果表明：目标表面等效电流分布以及分段等效合成孔径张角会随入射角度发生改变,其中分段等效合成孔径张角会影响方位向分辨率;双基地雷达成像结果中包含更丰富的电磁散射信息。本研究可为SAR系统设计和验证、典型目标回波特性数据收集以及如何基于高解析度SAR影像进行目标识别等研究提供借鉴与参考。     

适用于RPA工具的卷积序列推荐算法

在机器人流程自动化(Robotic Process Automation, RPA)软件中,经常采用序列推荐系统让机器人完成判断、选择等人工处理的任务。然而常用的序列推荐系统受限于序列信息的提取困难等问题,难以得到广泛应用。为了解决这一问题,构建一种基于Inception的卷积序列推荐模型,把时间和潜在空间中的用户行为序列信息嵌入进一幅“图像”中,并通过动态和静态2种不同的卷积层提取其中的局部特征,全面地提取用户的短期兴趣偏好,同时将用户嵌入矩阵作为用户的长期兴趣偏好嵌入到卷积层的输出中,共同构建完整的用户兴趣偏好,提升推荐性能。通过在3种公开数据集MovieLens 1M、Gowalla、Steam上分别进行实验,验证了基于Inception的卷积序列推荐模型的性能优于最新的序列推荐模型,在Top-N序列推荐的3种评价指标中（精确率、召回率、平均AP值）,平均提升幅度在10%左右,单个指标上的最大提升幅度为14%。     

Evolutionary digital twin: A new approach for intelligent industrial product development

To fulfill increasingly difficult and demanding tasks in the ever-changing complex world, intelligent industrial products are to be developed with higher flexibility and adaptability. Digital twin (DT) brings about a possible means, due to its ability to provide candidate behavior adjustments based on received “feedbacks” from its physical part. However, such candidate adjustments are deterministic, and thus lack of flexibility and adaptability. To address such problem, in this paper an extended concept – evolutionary digital twin (EDT) and an EDT-based new mode for intelligent industrial product development has been proposed. With our proposed EDT, a more precise approximated model of the physical world could be established through supervised learning, based on which the collaborative exploration for optimal policies via parallel simulation in multiple cyberspaces could be performed through reinforcement learning. Hence, more flexibility and adaptability could be brought to industrial products through machine learning (such as supervised learning and reinforcement learning) based self-evolution. As a primary verification of the effectiveness of our proposed approach, a case study has been carried out. The experimental results have well confirmed the effectiveness of our EDT based development mode.     

DEAR-MULSEMEDIA: Dataset for emotion analysis and recognition in response to multiple sensorial media

Traditionally, emotion recognition is performed in response to stimuli that engage either one (vision: image or hearing: audio) or two (vision and hearing: video) human senses. An immersive environment can be generated by engaging more than two human senses while interacting with multimedia content and is known as MULtiple SEnsorial media (mulsemedia). This study aims to create a new dataset of multimodal physiological signals to recognize emotions in response to such content. To this end, four multimedia clips are selected and synchronized with fan, heater, olfaction dispenser, and haptic vest to augment cold air, hot air, olfaction, and haptic effects respectively. Furthermore, physiological responses including electroencephalography (EEG), galvanic skin response (GSR), and photoplethysmography (PPG) are observed to analyze human emotional responses while experiencing mulsemedia content. A t-test applied using arousal and valence scores show that engaging more than two human senses evokes significantly different emotions. Statistical tests on EEG, GSR, and PPG responses also show a significant difference between multimedia and mulsemedia content. Classification accuracy of 85.18% and 76.54% is achieved for valence and arousal, respectively, using K-nearest neighbor classifier and feature-level fusion strategy.     

基于NOMA的5G超密网计算迁移与资源分配策略

针对5G超密网中移动设备计算能力不足、频谱资源有限的问题,提出了一种基于非正交多址接入（NOMA）的计算迁移与带宽分配策略。首先,对系统模型进行了分析,并在此基础上以最小化设备计算代价为目标对所研究的问题进行形式化定义;然后,将该问题分解成设备的计算迁移、系统的带宽分配和设备的分组匹配三个子问题,并利用模拟退火、内点法和贪心算法对这三个子问题进行求解;最后,通过联合优化算法对上述子问题进行交替性迭代求解,最终获得最优计算迁移和资源分配策略。仿真结果表明,所提出的联合优化策略不但优于传统的正交多址接入（OMA）方式,而且能获得比平均分配带宽的NOMA技术更低的设备计算代价。     

装配式建筑自上而下设计信息协同与模型构建

为顺应国家建筑产业化、智能化发展的政策引导,针对当前装配式建筑专业信息缺乏关联、模 型利用不充分、数据传递效率低下等问题,提出基于建筑信息模型(BIM)技术的自上而下设计方法。以装配式 建筑的设计阶段为切入点,详细阐述了基于 BIM 技术的装配式建筑自上而下设计流程。结合装配式建筑的标 准化构件设计和组装特点,提出一种基于自上而下设计的装配模型,并从层级和专业 2 个角度描述了模型的架 构。针对各专业模型传递时的数据信息标准化、一致性问题,通过工业基础类(IFC)标准对装配模型进行实体和 属性集的扩展,提出基于 IFC 标准的各专业、各阶段的自上而下设计信息协同。经实例验证了 IFC 扩展装配模 型的自上而下设计可行性,为装配式建筑的协同设计提供了有益的方法借鉴。     

虚拟现实全景流体绘画系统的可用性研究

虚拟现实环境给三维空间场景应用带来更强的真实感和身临其境的体验。在虚拟现实场景中经 常需要使用 360全景图像生成沉浸式环境背景,其中全景流体动力学绘画系统提供了一个新的全景图绘制方 法。在调研虚拟现实技术在流体绘画领域应用的适用性测试中,本文综合运用用户测试法、操作日志分析法、 事后问卷调查法和访谈法,对基于虚拟现实的全景式流体绘画创作系统进行可用性评估,并创意性地使用可视 化方法辅助日志分析。所有的受试者均认为系统的可用性不低于一般水平,利用该系统完成流体绘画的创作工 作不需要受试者具有专业绘画经验。受试者对系统的功能效果持积极肯定的意见,并期待尝试用其他交互形式 代替手柄类交互方式以促进艺术创作的表达。其研究方法和结果将为同类系统的开发和评估提供参考。     

New empirical model to evaluate groundwater flow into circular tunnel using multiple regression analysis

There are various analytical, empirical and numerical methods to calculate groundwater inflow into tunnels excavated in rocky media. Analytical methods have been widely applied in prediction of groundwater inflow to tunnels due to their simplicity and practical base theory. Investigations show that the real amount of water infiltrating into jointed tunnels is much less than calculated amount using analytical methods and obtained results are very dependent on tunnel‘s geometry and environmental situations.In this study, using multiple regression analysis, a new empirical model for estimation of groundwater seepage into circular tunnels was introduced. Our data was acquired from field surveys and laboratory analysis of core samples. New regression variables were defined after perusing single and two variables relationship between groundwater seepage and other variables. Finally, an appropriate model for estimation of leakage was obtained using the stepwise algorithm. Statistics like R, R~2, R_P~2 and the histogram of residual values in the model represent a good reputation and fitness for this model to estimate the groundwater seepage into tunnels. The new experimental model was used for the test data and results were satisfactory. Therefore, multiple regression analysis is an effective and efficient way to estimate the groundwater seepage into tunnels.     

Unanticipated multiple seam stresses from pillar systems behaving as pseudo gob–case histories

Underground coal mining in the U.S. is conducted in numerous regions where previous workings exist above and/or below an actively mined seam. Miners know that overlying or underlying fully extracted coal areas, also known as gob regions, can result in abutment stresses that affect the active mining. If there was no full extraction, and the past mining consists entirely of intact pillars, the stresses on the active seam are usually minimal. However, experience has shown that in some situations there has been sufficient yielding in overlying or underlying pillar systems to cause stress transfer to the adjoining larger pillars or barriers, which in turn, transfer significant stresses onto the workings of the active seam. In other words, the overlying or underlying pillar system behaves as a ‘‘pseudo gob." The presence of a pseudo gob is often unexpected, and the consequences can be severe. This paper presents several case histories, summarized briefly below, that illustrate pseudo gob phenomenon:(1) pillar rib degradation at a West Virginia mine at 335 m depth that contributed to a rib roll fatality,(2) pillar rib deterioration at a Western Kentucky mine at 175 m depth that required pillar size adjustment and installation of supplemental bolting,(3) roof deterioration at an eastern Kentucky mine at 400 m depth that stopped mine advance and required redirecting the section development,(4) coal burst on development at an eastern Kentucky mine at 520 m depth that had no nearby pillar recovery, and(5) coal burst on development at a West Virginia mine at the relatively shallow depth of 335 m that also had no nearby pillar recovery. The paper provides guidance so that when an operation encounters a potential pseudo gob stress interaction the hazard can be mitigated based on an understanding of the mechanism encountered.