应用SVM模型对某县级水厂水源水指标的预测

目的使用湖北省某县级自来水厂日常监测化学耗氧量数据进行浓度预测。方法运用METLAB应用软件构建水质预测模型,并对湖北省枝江市水厂水源水化学耗氧量浓度和浊度进行预测。结果建模后预测结果显示,水源水COD周均和月均数据相关系数较大,平均相对误差<10.00%,模型拟合性能较好;浊度周均和月均数据建模相关系数较低,平均相对误差>16.00%,模型拟合性能较差。结论该模型可以用于预测COD浓度,周均效果优于月均,预测浊度效果较差,需进一步考虑其他算法进行预测。     

使用最小二乘支持向量机技术预测传染病发病率的研究

目的将最小二乘支持向量机(LS-SVM)技术应用到传染病预测中,寻找更加理想的预测结果。方法以某市1991—2002年乙型肝炎(乙肝)月发病率数据建立最小二乘支持向量机预测模型,对2003年1—6月的月发病率进行预测。结果 IS-SVM预测值分别为0.709 9,0.668 1,0.502 5,0.685 1、0.578 5,0.773 7,通过与径向基函数(RBF)神经网络模型和累积式自回归动平均模型(ARIMA)预测结果进行比较,预测精度明显高于RBF网络模型和ARIMA模型,相对误差明显减少,仅为ARIMA模型的23.62%,RBF网络模型的54.69%。结论 LS-SVM模型对乙肝发病率的预测精度更高,效果更好,也验证了支持向量机方法预测能力出色的理论优点,证明了支持向量机技术在传染病预测领域同样有着良好的表现。     

支持向量机原理及其在医学分类中的应用

目的 介绍一种具有较高精度的分类模型--支持向量机在解决分类问题时的应用.方法 以胃癌流行病学调查资料为例分别建立支持向量机、决策树、logistic 回归模型,比较三种模型性能优劣.结果 对于测试集样本SVM预测精度为99.052%,C5.0决策树预测精度为93.365%,logistic 回归预测精度为90.995%,SVM具有良好的泛化能力.结论 当传统统计分析条件不能得到满足或效果不佳时支持向量机能够达到良好的预测结果,在医学领域具有较好的应用前景.     

基于支持向量机的麻疹发病率预测研究

目的 建立一种基于支持向量机的麻疹发病率预测模型,为麻疹的预防控制决策提供参考依据。方法 收集我国麻疹1996-2015年发病率数据,以1996-2014年发病率数据为训练样本,以2015年发病率数据为检验样本,采用支持向量机回归算法建立我国麻疹年发病率的预测模型,并对我国2016-2018年的麻疹发病率进行预测。结果 经该模型的预测,麻疹发病率的预测值和实际值吻合度较高,平均相对误差为0.620 07%。我国2016-2018年麻疹的发病率分别为3.23/10万、3.13/10万、3.79/10万。结论 采用支持向量机回归模型预测我国麻疹的年发病率是可行有效的。     

机器学习算法在预测男男性行为人群中HIV感染的应用

目的应用不同机器学习算法预测男男性行为(MSM)人群HIV感染状况的比较。方法将四种机器学习算法(logistic回归、神经网络、随机森林和支持向量机)的预测结果和实验室的检测结果相比较,分类性能的评价指标采用ROC曲线下面积(AUC)、灵敏度、特异度和准确度(PRE)。结果四种算法在训练集和测试集上均具有较理想的分类效能,训练集的结果略好于测试集。和logistic回归分析相比较,其他几种算法均提高了分类预测效能:神经网络提高18.4%(AUC:0.909,95%CI:0.903～0.915),随机森林提高19.7%(AUC:0.922,95%CI:0.920～0.924),支持向量机提高22.3%(AUC:0.948,95%CI:0.947～0.949)。其中支持向量机的分类性能最好,分类的灵敏度为97.5%,特异度为99.1%,准确度(PRE)为98.9%。结论机器学习算法显著地提高了MSM人群中HIV感染的预测效能,可以较准确地识别MSM人群中HIV感染者与未感染者,为及时地提供预防与治疗服务提供了依据,同时避免了医疗资源的浪费。     

支持向量机和Logistic回归在MSM人群艾滋病患病风险预测中的应用北大核心

目的探讨两种不同机器学习算法在海口市男男性行为(MSM)人群HIV患病情况风险预测中的应用。方法收集经过数据清洗和筛选后共计3184个样本数据作为抽样集,生成随机序列抽取编号,保证数据均衡的情况下,随机抽样200个,分别运用支持向量机和Logistic回归两种机器算法进行预测分类,依据4个评价指标,比较两种算法预测结果的准确度。并在抽样集上,采用Logistic回归模型对HIV患病情况进行影响因素分析。结果支持向量机灵敏度为0.955,特异度为0.938,准确度为0.950,相对提升度为158.333%;Logistic回归的灵敏度为0.545,特异度为0.750,准确度为0.600,相对提升度为63.158%;哨点监测数据中,Logistic回归模型采用逐步回归法,筛选出3个HIV患病影响因素,采取安全措施和接受预防服务为保护因素,OR值为0.410和0.820;患有伴随性病为危险因素,OR值为3.626。结论在分类与预测方面,支持向量机的结果比Logistic回归模型更加精确。MSM人群应当注意采用安全措施、接受干预服务,提高预防性病的风险意识,减少患HIV的可能性。     

支持向量机在洪灾区创伤性应激障碍预测中的应用

应用支持向量机对洪灾区居民创伤性应激障碍(PTSD)的发生进行预测.使用美国<精神障碍的诊断统计手册>第四版(DSM-IV)中关于PTSD的诊断标准对洪灾区成年人进行评定,以是否发生PTSD为应变量,以影响PTSD发生的23个因素为自变量,建立基于支持向量机(SVM)的预测模型,对遭受洪灾后PTSD的发生进行预测.将影响PTSD发生的23个因素纳入预测模型后,测试集SVM分类与实际类别的一致率为88.05%,灵敏度为75.0%,特异度为89.4%.结论:应用SVM建立预测模型对于洪灾区PTSD发生的预测具有较好的效果,被纳入的23个因素作为输入向量有良好的预测效率.     

支持向量机和Logistic回归在MSM人群艾滋病患病风险预测中的应用

目的探讨两种不同机器学习算法在海口市男男性行为(MSM)人群HIV患病情况风险预测中的应用。方法收集经过数据清洗和筛选后共计3184个样本数据作为抽样集,生成随机序列抽取编号,保证数据均衡的情况下,随机抽样200个,分别运用支持向量机和Logistic回归两种机器算法进行预测分类,依据4个评价指标,比较两种算法预测结果的准确度。并在抽样集上,采用Logistic回归模型对HIV患病情况进行影响因素分析。结果支持向量机灵敏度为0.955,特异度为0.938,准确度为0.950,相对提升度为158.333%;Logistic回归的灵敏度为0.545,特异度为0.750,准确度为0.600,相对提升度为63.158%;哨点监测数据中,Logistic回归模型采用逐步回归法,筛选出3个HIV患病影响因素,采取安全措施和接受预防服务为保护因素,OR值为0.410和0.820;患有伴随性病为危险因素,OR值为3.626。结论在分类与预测方面,支持向量机的结果比Logistic回归模型更加精确。MSM人群应当注意采用安全措施、接受干预服务,提高预防性病的风险意识,减少患HIV的可能性。     

基于支持向量机的缺血性脑卒中患病风险预测

目的分析缺血性脑卒中发病因素,建立缺血性脑卒中患病风险预测模型,为脑卒中的早期筛查与防治提供依据。方法应用统计学方法筛选与缺血性脑卒中患病相关的指标,通过支持向量机算法构建预测模型,并对模型进行评估。结果采用支持向量机算法构建的简单个人和复杂临床水平模型的训练集准确率分别为97.48%、97.09%,测试集准确率分别为91.47%、91.09%。结论支持向量机算法构建的风险预测模型,具有很好的时效性与准确性,对脑卒中早期筛查和干预具有一定的应用价值。     

重采样技术在中老年居民糖尿病不平衡数据分类中的应用

目的 利用重采样技术提高我国中老年居民糖尿病不平衡数据的分类预测效果。方法 采用随机欠采样、随机过采样、合成少数类过采样(synthetic minority oversampling technique, SMOTE)以及自适应合成抽样(adaptive synthetic sampling, ADASYN)四种重采样技术处理CHARLS数据库中糖尿病不平衡数据,比较重采样前后logistic回归、支持向量机、随机森林的分类性能,采用G-means和AUC评价模型的预测效果。结果 对CHARLS糖尿病不平衡数据集,logistic回归、支持向量机、随机森林模型的G-means分别为0.222 7、0、0,AUC分别为0.761 2、0.736 3、0.742 9,logistic回归模型显著优于支持向量机,模型准确率(χ²=1 231.501,P<0.001)及AUC值(Z=2.634, P=0.028)的差异均具有统计学意义。四种重采样技术处理后模型的G-means均提高,特别是SMOTE和ADASYN技术;此外,随机欠采样不能显著提高logistic回...     

Calcium intake in the elderly: validation of a dietary questionnaire

Epidemiological studies of calcium and osteoporosis have been hampered by the lack of a suitable tool for assessing calcium intake. This report describes a new frequency and amount questionnaire for measuring present and past calcium intake in the elderly. The validity of the questionnaire was tested against two commonly used standards of dietary assessment, five-day duplicate diets and seven-day weighed dietary inventories. The resulting correlation coefficients were, respectively, r = 0.76 and r = 0.69, while that for repeatability was r = 0.84. Furthermore, the questionnaire categorized subjects into thirds of the distribution of intake with almost no gross misclassification. It is suggested that the present findings may be extended to the majority of normal, healthy elderly subjects, implying wide application for the questionnaire in the assessment of calcium intake in the elderly.     

Self-reported health problems among Swedish miners one year after unemployment

Unemployment is considered to be a public health concern sincedeterioration in the health of the unemployed is often anticipated.However, for some groups, such as miners, unemployment mightimprove health due to a cessation of potentially harmful occupationalexposures. This study evaluates the health of 79 miners in oneSwedish iron-ore mine, and 226 age-matched controls from thegeneral population, during one year after the closure of themine. The participants received a questionnaire regarding medicalhistory and subjective symptoms at the beginning of the studyperiod, and after one year. Statistically significant negativeeffects on self-reported health attributable to unemploymentwere not found, although neuropsychiatric symptoms were morecommon among the unemployed miners. The miners reported a statisticallysignificant improvement in grip force (p=0.031). They had asignificantly higher prevalence of symptoms associated withmining related exposures when compared with the population controls;pain in the upper extremities [relative risk (RR)=2.27, 95%confidence interval (Cl)=1.44–3.59), back pain (RR=1.84;Cl=1.237–2.75), vasospastic disease of the fingers (RR=2.05;Cl=1.18–3.57) and obstructive respiratory symptoms (attacksof dyspnea and wheezing: RR=3.67; Cl=1.167–11.6).     

Excès de cas humains et animaux de tularémie en France au cours de l’hiver 2007–2008 : émergence ou phénomène isolé ?

Context

Tularemia is a zoonosis affecting humans and hares in France. We describe the results of surveillance in both species, in 2007 and 2008.

Methods

Human tularemia cases are mandatorily notifiable in France since 2003. In hares, surveillance relies on volunteer hunter associations in all districts of the country. Data from mandatory reports and volunteer surveillance in 2007/2008 were analyzed and compared with previous results.

Results

In 2007/2008, 144 cases were reported in humans and 117 cases in hares. This was a 100% increase compared to previous years. Human cases differed from those of previous years only by the frequency of contact with breeding animals. Human cases without any documented risk exposure were also more frequent.

Conclusion

An increase of tularemia cases occurred in 2007/2008 in both species. Complementary studies are needed to identify the species reservoir in France to understand the causes of this peak of cases.     

Occupational health hazards in mining: an overview

This review article outlines the physical, chemical, biological, ergonomic and psychosocial occupational health hazards of mining and associated metallurgical processes. Mining remains an important industrial sector in many parts of the world and although substantial progress has been made in the control of occupational health hazards, there remains room for further risk reduction. This applies particularly to traumatic injury hazards, ergonomic hazards and noise. Vigilance is also required to ensure exposures to coal dust and crystalline silica remain effectively controlled.     

Manufacture and use of nanomaterials: current status in the UK and global trends

This paper provides an overview of the production and use of nanomaterials (NMs), particularly in the UK. Currently, relatively few companies in the UK are identifiable as NM manufacturers, the main emphasis being the bulk markets in metals and metal oxides, and some niche markets such as carbon nanotubes and quantum dots. NM manufacturing in the UK does not reflect the global emphasis on fullerenes, nanotubes and fibres. Some assumptions have been made about the types of NM that are likely to be imported into the UK, which currently include fullerenes, modified fullerenes and other carbon-based NMs including nanotubes. Many university departments, spin-offs and private companies have developed processes for the manufacture of NMs but may only be producing small quantities for research and development (R&D) purposes. However, some have the potential to scale up to produce large quantities. The nanotechnology industry in the UK has strong R&D backup from universities and related institutions. This review has covered R&D trends at such institutions, and appropriate information has been added to a searchable database. While several companies are including NMs in their products, only a few (e.g. manufacturers of paints, coatings, cosmetics, catalysts, polymer composites) are using nanoparticles (NPs) in any significant quantities. However, this situation is likely to change rapidly. There is a need to collect more information about exposure to NPs in both manufacturing and user scenarios. As the market grows, and as manufacturers switch from the micro- to the nanoscale, the potential for exposure will increase. More research is required to quantify any risks to workers and consumers.     

深圳公立医院管理体制改革制度设计

深圳公立医院管理体制改革实行政事分开、管办分开,在理事会架构下按法定机构模式组建市医管中心,落实公立医院运营管理自主权.作者从当前公立医院管理体制的弊端入手,介绍了深圳市进行公立医院管理体制改革的基本思路及改革方案设计的主要举措,深入剖析了的改革方案的特点,并对改革效果进行了预测.     

Effects of nitrogen dioxide on red blood cells of rats: Changes in components of red cell membranes during in vivo exposure to NO2

Red cell membranes, prepared from red blood cells of rats exposed to 4, 10, or 20 ppm nitrogen dioxide (NO₂) for 1 to 10 days, were examined for evidence of changes in membrane components. Appreciable changes were not found in contents of phospholipid and cholesterol during exposure to 10 ppm NO₂. By contrast, protein content altered with the time of exposure. Moreover, changes in protein composition were observed by employing sodium dodecyl sulfate — polyacrylamide gel electrophoresis. Twenty-four-hour exposure to NO₂ at the concentration above 10 ppm resulted in a marked increase in the percentage of lysophosphatidylethanolamine (LysoPE) to the total phospholipids. The prolonged exposure to 10 ppm NO₂ gave rise to a further increase in LysoPE, whereas the percentage of phosphatidylethanolamine (PE) showed a gradual decrease. A 1-day exposure to 4.0 ppm NO₂ also caused an increase in sialic acid content and decreases in those of PE and hexose. In addition to contents of these components the percentage of LysoPE increased 5 days after exposure and the elevated values were maintained up to the end of exposure period. These results demonstrate that red blood cells in circulation exhibit different membrane properties in terms of lipid and carbohydrate composition during 10 days of exposure to 4.0 ppm NO₂.     

Investigating clusters in the workplace and beyond

Clusters of disease are common and occur in the workplace and in the general community. They often arouse considerable concern among the population. Investigations have sometimes lead to exciting new knowledge, but in general the investigation of clusters is difficult and often unrewarding, especially for community clusters. In the workplace, investigations are more likely to find associations and even new causes, but still many clusters remain enigmatic. Despite this, there are many reasons for investigating clusters, including allaying community concern and identifying uncontrolled exposures. A structure for investigating clusters in the workplace is suggested.