排序方式: 共有35条查询结果,搜索用时 31 毫秒
1.
2.
3.
This study investigated the effects of sampling conditions on composition of VOCs including different flow restrictors, SUMMA volumes, sampling heights and wind speed. Results showed that under the six sampling heights, the concentrations of main species of VOCs were slightly different, while wind speed has a greater impact on the content of VOCs source profiles. With the increase in wind speed, the weighted percentage of high-carbon aromatic hydrocarbons was higher. In addition, there were an extremely different profile between the normal production and shutdown conditions of delayed coking unit. In order to compare the emission characteristics of VOCs in process units from S and C refineries, the samples were collected in catalytic cracking unit, continuous catalytic reforming unit, and delayed coking unit. C3-C5 alkanes and low-carbon aromatic hydrocarbons were the main components in continuous catalytic reforming unit, about 67.1% and 34.9%, respectively. For delayed coking unit, the total content of high carbon alkane C6-C12 was significantly higher than other units, about 30.5% and 24.4%, respectively. The weighted percentage of C2-C5 low-carbon alkane were abundant, and propylene was higher in catalytic cracking unit. The emission characteristics obtained were consistent with the processing technology of production units. The results indicated that the source profiles from the same production unit in different refineries have similarities, but there were also significant differences. The emission characteristics of VOCs in the device could provide data support for source apportionment work in the production units. 相似文献
4.
6.
随着人民生活水平的不断提高,肿瘤疾病的人数在不断增多,其中肺癌是21世纪严重危害人类健康的重大疾病.为此提出一种基于电子病历的肺癌诊断决策树方法.首先分析肺癌电子病历的特点以及决策树存在结构不稳定、过拟合等现象,运用主成分分析法结合C5.0算法构建的优化决策树模型.首先,建立主成分特征根大于1以及主成分累计贡献率大于85%的特征降维两种方法,然后通过C5.0算法建立决策树模型和剪枝操作,最后给出数据预处理过程及模型的执行流程和测试结果.实验结果分析,改进的算法有较好的准确率以及良好的可扩展性,从而验证了改进后的算法对于辅助肺癌临床实验具有重要的意义. 相似文献
7.
集成学习的一个重要目标是获得一组差异性大的基分类器来构建集成分类器。为实现这一目标,提出一种基于抽样与约简的集成学习算法ELSR。该算法采用多模态扰动策略来训练基分类器。首先,采用多次抽样策略从训练集中抽样产生k个抽样集;其次,使用粗糙集的属性约简技术对每个抽样集进行约简;第三,在每一个约简之后的抽样集上分别训练一个基分类器;最后,利用一个验证集对每个基分类器进行性能测试,并根据测试结果选择一组合适的基分类器来构建集成分类器。在UCI数据集上的实验表明:当采用KNN算法或者C4.5算法来训练基分类器时,ELSR的分类性能总是要优于现有的集成学习算法。 相似文献
8.
医疗大数据的广泛应用,给心脑血管疾病的研究带来新的机遇.合理利用数据特征进行分析,对心脑血管疾病在医药作用、疾病关联、风险预警等方面的发展具有推动意义.但心脑血管疾病大数据存在一定的个性特征,特别是关联性强、周期性长、隐私性敏感等方面,给数据的应用造成一定的影响.文章阐述了医疗大数据的发展对心脑血管疾病研究的影响,并主要分析心脑血管疾病大数据所体现的特征及应用.同时结合医疗领域与信息领域,提出了心脑血管疾病大数据所面临的挑战和展望. 相似文献
9.
针对临床路径的复杂性和模糊不确定性,对其进行综合分析,计算出临床路径诊疗单元的平均治疗天数以及临床路径的总治疗天数,分别作为实验的样本数据和最终评价指标。在此基础上,利用BP神经网络的方法对临床路径进行优化建模。并通过实验对比分析隐层神经元数量对临床路径优化结果的影响,发现神经元数量增多时,错误率明显下降,迭代次数呈上下波动。最后选取7-9-1的BP神经网络结构,以某地区三甲医院的糖尿病加高血压临床路径为例,进行仿真实验。结果显示,训练模型的输出值与期望值之间的相对误差范围在0-0.2%之内,测试模型的输出值与期望值之间的相对误差范围在0-0.1%之内。说明该模型具有较强的自学习自适应能力,能够有效地优化临床路径,减少患者的治疗天数。 相似文献
10.
文章对培训过程中培训需求的预测、培训环境的营造、评估标准的设定等进行了深入分析,目的在于帮助企业培训部门、培训研究人员以及培训师设计出符合企业发展、提高员工绩效水平的培训项目。 相似文献