基于样本分析的试题难度系数修正算法

针对当前题库试题难度系数的确定和评价标准不明确,提出一种以考试结果为样本获得难度系数并自我修正的算法:将样本数据序列{dk}构造到希尔伯特空间上,利用空间的完备性,从理论上确保序列{dk}收敛于常量D。基于该算法构造机器学习模型,它具有快速收敛和对历史数据进行学习和修正的特点。     

流动注射在线预富集测定环境样品中痕量铅

流动注射在线无过滤沉淀-溶解系统被应用于火焰原子吸收测定环境水样中痕量铅。通过样品溶液和氨溶液的混合已经可以达到痕量铅的在线析出。结果沉淀物无需过滤就被收集在编结反应器的内壁上。引入1mol/L的HNO3溶液来溶解沉淀物并将分析物输送至火焰原子吸收光谱仪(FAAS)进行在线分析。样品载入流速为3.5mol?L-1/m in,富集时间40 s,增强因子37,检测频率60/h。检测限(3σ)为7.5μg/L。50μg/L水平时,精密度(RSD,n=11)为2.9%。采用该方法测定用简单水溶液标准校准后的国家标准材料(GBW 08607,河道水),所得铅浓度0.96±0.04g/g。结果与标准值(1.00±0.02g/g)非常接近。该方法也可以成功应用于多种水样中痕量铅的测定。     

管道漏磁检测缺陷识别技术

针对神经网络法对管道缺陷进行识别存在所需样本数据量大、容易陷入局部极值、泛化能力没有保证等问题,提出一种新的机器学习方法,即支持向量机法.介绍了支持向量机的分析原理,以碗状缺陷为例,采用ANSYS有限元分析软件对40组不同尺寸的缺陷进行仿真,将得到的40组漏磁数据和10组实测数据作为学习样本,另外10组新数据作为验证数据,以MATLAB软件为平台进行了识别实验.实验结果表明,识别误差均在5％以下,且泛化能力强,识别方法简单,在有限样本情况下比神经网络法更具优势.     

Fuzzing技术的样本数据组合研究*

提出了基于样本数据组合的Fuzzing技术,并抽象出了解决样本数据覆盖问题（data sample covering problem,DSCP）的数学模型;为了更好地解决样本数据覆盖问题,提出了改进遗传算法(developed genetic algorithm,DGA),通过实例实验说明了DGA的有效性,并且通过仿真实验,验证了算法在求解复杂协议文件样本数据覆盖问题时比贪心算法和简单遗传算法具有更高的效率。     

用PSO方法搜索基于MLE的ARMA模型参数

针对小样本条件下用矩估计(ME)方法获取ARMA模型参数粗略的缺点,将粒子群优化算法(PSO)用于小样本ARMA模型参数的极大似然估计(MLE),以获得概率上最优的数字解。在分析基于ARMA模型似然函数的基础上,详细分析了PSO的思想、方法和评价指标。以实际例证显示了联合PSO优化方法估计AMAR模型参数的优良特性,并从算法和似然函数角度分别阐释了形成利弊的原因。     

(<Emphasis Type="Italic">Z,Z</Emphasis>)-6,9-Heneicosadien-11-One: Major Sex Pheromone Component Of Painted Apple Moth, <Emphasis Type="Italic">Teia anartoides</Emphasis>

(Z, Z)-6,9-Heneicosadien-11-one (Z6Z9-11-one-21Hy) was identified as the major sex pheromone component of the painted apple moth (PAM), Teia anartoides (Lepidoptera: Lymantriidae), on the basis of (1) comparative gas chromatographic-electroantennographic detection (GC-EAD) analyses, GC-mass spectrometry (MS), high-performance liquid chromatography (HPLC)-MS, and HPLC-UV/visible spectroscopy of pheromone gland extracts and authentic standards; (2) GC-EAD analyses of effluvia of calling females; and (3) wind tunnel and field trapping experiments with a synthetic standard. In field experiments in Australia, synthetic Z6Z9-11-one-21Hy as a single component attracted male moths. Wind tunnel experiments suggested that a 4-component blend consisting of Z6Z9-11-one-21Hy, (6Z,9R,10S)-cis-9,10-epoxy-heneicosene (Z6-9R10S-epo-21Hy), (E, E)-7,9-heneicosadien-6,11-dione (E7E9-6,11-dione-21Hy), and 6-hydroxy-(E, E)-7,9-heneicosadien-11-one (E7E9-6-ol-11-one-21Hy) (all present in pheromone gland extracts) might induce more males to orient toward, approach, and contact the source than did Z6Z9-11-one-21Hy as a single component. Additional experiments are needed to determine conclusively whether or not Z6-9R10S-epo-21Hy, E7E9-6,11-dione-21Hy, and E7E9-6-ol-11-one-21Hy might be minor sex pheromone components of PAM. Moreover, attractiveness of synthetic pheromone and virgin PAM females needs to be compared to determine whether synthetic pheromone could replace PAM females as trap baits in the program to monitor eradication of exotic PAM in New Zealand.     

基于组合预测模型的小样本轴承故障分类诊断

滚动轴承是旋转机械内常出现问题的重要部件,其故障情况复杂且难以诊断.基于小样本故障数据学习环境,针对小样本学习在提取真实特征值与目标特征值时有较大差异且泛化能力较弱的问题,提出一种采用半监督变分自编码器与LightGBM分类模型相结合的小样本学习模型LSVAE,并利用基于高斯过程的贝叶斯优化改进算法对LightGBM的超参数进行了优化处理,有效地解决了小样本学习性能不稳定,提取特征能力弱,过拟合等问题,并在凯斯西储大学发布的轴承实验数据集上进行了对比实验,结果表明LSVAE模型在面向小样本数据空间时有着更优的诊断准确率.     

基于AERF模型的油井结蜡预测

油井结蜡是一种在开发以及开采油田时对油井正常产出造成了负面影响的现象,该现象会引起油流通道堵塞,导致油井开采过程中出油量降低.对油井结蜡状况做出智能预警,完成油井设备提前修复,对油田提高产能效率、降低维护成本及智能化管理有非常关键的价值.为了解决油井正常数据和结蜡数据严重不平衡问题,本文引入了自适应合成抽样法(ADASYN)和最近邻规则欠抽样法(ENN)两种非均衡样本处理方法,分别对类别为结蜡的样本和非结蜡的样本进行处理,最终使用随机森林算法对新构成的数据集训练,构造出AERF智能模型来预测油井结蜡.实验结果表明,提出的AERF模型在油井的结蜡数据集中预测效果更佳,明显地提高了预测精度.     

多代表点的加权近邻分类算法

传统的KNN算法存在分类效率低等缺点.针对这些缺点,本文提出一种高效的结合多代表点思想的加权KNN算法,利用变精度粗糙集上下近似区域的概念,结合聚类算法生成代表点集合构造分类模型,再运用结构风险最小化理论优化分类模型并对影响分类模型的因素进行分析.分类过程中根据测试样本与各代表点的相似度,得到测试样本的相对位置.其中属于样本点下近似区域的测试样本可直接判断其类别.若测试样本在其他区域,则根据测试样本与各代表点的相对位置对各代表点覆盖范围内的样本进行加权后判断测试样本的类别.在文本分类领域的数据集上进行实验,结果表明该算法能有效的提高分类模型的性能.     

A small samples training framework for deep Learning-based automatic information extraction: Case study of construction accident news reports analysis

Knowledge management is crucial for construction safety management. Widely collected and well-organized safety-related documents are recognized to be significant in raising the workers' security awareness and then to prevent hazards and accidents. To improve document processing efficiency, automatic information extraction plays an important role. However, currently, automatic information extraction modeling requires large scale training datasets. It is a big challenge for the engineering industry, especially for the fields which heavily rely on the experts’ knowledge. Limited data sources, and high time and labor costs make it not practical to establish a large-scale dataset. This work proposed a natural language data augmentation-based small samples training framework for automatic information extraction modeling. With the designed cross combination-based text data augmentation algorithm, the deep neural network can be employed to build up automatic information extraction models without large-scale raw data and manual annotations. Characters semantic coding is employed to avoid word segmentation and make sure that the framework can be utilized in different writing language systems. The BiLSTM-CRF model is adopted as the detection core to conduct character classification. Through a case study of two independent accident news report datasets analysis, the proposed framework has been validated. A reliable and robust automatic information extraction model can be established, even though with small samples training.