一种扩展的朴素贝叶斯分类器改进算法

文中研究贝叶斯分类器家族中的一种扩展朴素贝叶斯分类器。此种扩展朴素贝叶斯分类器满足两个条件:一是类结点是所有属性的父结点;二是每个属性最多有一个属性父结点。其中有代表性的两种算法是贪婪爬山算法(Hill Climb-ing Search,即HCS算法)和超父结点算法(Superparent,即SP算法)。对两种算法进行了分析和比较,并在此基础上提出了一种改进算法。通过实验验证所改进的分类器是正确的、有效的。     

Heat stress and bulkiness of chemical protective clothing impair performance of medical personnel in basic lifesaving tasks

The study examined the impact of chemical protective (CP) clothing on the performance of lifesaving tasks in thermoneutral and cold conditions. Eleven males performed pre-exercise followed by lifesaving tasks wearing either field combat uniform at 21°C (U) or CP clothing at 21°C (CPN) and–;5°C (CPC). The tasks were ventilating a doll (VA) and connecting an intravenous line (IV). Mean skin temperature was significantly higher for CPN compared to U and CPC during pre-exercise, VA and IV. Changes in blood pressure were significantly greater with CP clothing than without during VA and IV. The number of breaths per min (in VA) and time needed for IV increased by 19% (p < 0.05) and 18%, respectively, for CPN compared to U. Due to the cold, the additional increment was 5% and 17%, respectively, for CPC. Wearing of CP clothing in thermoneutral or in cold conditions may not prevent but, especially in the cold, significantly impede the performance of basic medical tasks. The findings of this study showed that performing medical tasks while wearing nuclear, biological and chemical protective clothing is impaired due to significant changes in physiological strain. This suggests that realistic training in local conditions as well as in cold conditions is needed to realise the restrictions due to protective clothing.     

基于树选择的TAN分类器

通过对朴素贝叶斯（NBC）分类器与传统的基于树扩展的贝叶斯（TAN）分类器的分析,对TAN分类器进行改进,提出CTAN分类器。朴素贝叶斯分类器对非类属性独立性进行完全独立假设,传统TAN则弱化所有属性的独立性．提出的CTAN则是通过操作TAN保留对数对部分相关属性有选择的进行弱化。CTAN改进的方向主要是对属性关系树的部分利用,通过实验证明,CTAN要优于传统TAN分类器。     

中国快递业发展的区域差异及动态演化

提出一种基于朴素贝叶斯分类法(naive Bayesian classifier,NBC)的城市快递人员辨识方法. 首先,通过相关问卷调查,研究快递派送人员的手机信令发生规则. 然后,依据北京市移动用户手机通信信令数据,利用问卷调查数据和手机信令数据2种数据源中同时包含的通信数据属性,建立通信数据与调查数据中类别变量(快递人员/非快递人员)之间的贝叶斯概率关系,以此为基础构建NBC模型并对其进行训练. 最后,使用未参与训练的样本数据测试标定后模型的准确性,测试结果显示快递人员的预测成功率达到88.3%. 结果表明:该方法具有较高的精度,可以满足实际应用需求.

     

Seismic fragility assessment of revised MRT buildings considering typical construction changes

The present study investigates the vulnerability assessment of the prototype revised Mandatory Rule of Thumb (MRT) buildings initially designed and detailed for three storeys bare frame building; later modified through variable number of storeys (three, four, and five) and different arrangement of infill walls (bare frame, soft-storey, irregular infilled, and fully infilled). The application of infill walls increases the fundamental frequencies, stiffness, and maximum strength capacity, but reduces the deformation capability than the bare frame building. The vulnerability was also reduced through infill walls, where the probability of exceeding partial-collapse and collapse damage reduced by 80% and 50%, respectively. Furthermore, the increased in storeys (three to five) also increases the failure probability, such that partial-collapse and collapse for fully infilled increases by almost 55% and 80%, respectively. All obtained results and discussions concluded that the structural sections and details assigned for MRT building is not sufficient if considered as bare frame and soft-storey. And increase in number of storeys causes building highly vulnerable although the infill walls were considered.     

钇基重稀土对NBC合金性能的影响

研究了钇基重稀土对NBC合金性能的影响。结果表明：NBC合金经过微量CU-3型稀土的微合金化处理后，其显微硬度、耐磨性和耐腐蚀性均有不同程度的提高，热膨胀系数有所下降；随着稀土加入量的增加，NBC合金的显微硬度逐渐提高，而其磨损量、平均腐蚀速率和热膨胀系数逐渐降低。     

一种基于动态贝叶斯网络的人体动作识别方法

针对人体动作识别过程中存在的效率及准确率问题,提出了一种基于混合贝叶斯网络模型的人体动作识别方法。通过Kinect采集人体动作RGB-D信息,采用OpenNi提取关节点信息并计算躯干角度,使用后验概率动态调整SVM分类器和朴素贝叶斯分类器权重,能够识别多种不同动作,使两个分类器互为补充,增加识别率。最后通过与单分类器的对比试验,验证了该算法具有更高的效率和识别率。     

朴素贝叶斯分类器的独立性假设研究

朴素贝叶斯分类器(NBC)是一种简洁而有效的分类模型。介绍了NBC模型的基本原理,并着重分析了该模型的独立性假设条件。在总结现有独立性假设研究的基础上,通过例子和实验分析得出结论:NBC模型的表现和独立性假设是否满足没有必然联系。     

硝化细菌纤维素溶液的流变性

以N,N-二甲基甲酰胺(DMF)为溶剂,按一定比例配制不同含氮量的硝化细菌纤维素(NBC)溶液。通过哈克(HAAKE)旋转流变仪,研究了溶液浓度、NBC含氮量对其溶液粘度的影响,以及溶液浓度与粘度的依赖性关系,并对NBC溶液的粘弹性能进行了应力扫描和频率扫描分析。结果表明,当NBC溶液浓度为0.5%~5.0%时,溶液体系为牛顿流体,粘度随着浓度的增加而增大,满足指数型增长关系;当体系浓度增大到5.5%时,NBC溶液为非牛顿流体,表现出剪切变稀效应,其剪切应力及剪切速率满足幂函数增长关系。随着NBC分子量的增加,溶液粘度增大。高分子量的NBC,有助于溶液体系稳定性的提高,同时也减弱了其对外界刺激性的响应。     

Nylon 6‐Polyesteramide Block Copolymers (NBC) and NBC/Phenolic Resin Composites, 1

Summary: A series of NBC/phenolic resin composites, containing 0, 1, 3, 5 or 7 wt.‐% of a powdered phenolic resin of different particle diameter, was prepared by the reaction injection molding (RIM) process. It was determined by SEM analysis that there exists a strong interaction between particles and matrix and that such interaction occurs through hydrogen‐type bonds as determined by FTIR analysis. According to the results it is thought that the glass transition temperature of the NBC/phenolic resin composites depends on two competing factors: the rigidity promoted by the hard solid filler and the flexibility imparted by the nylon 6 amorphous phase, whose proportion becomes more important with increasing amounts of phenolic resin particles. The elastic and flexural moduli of the NBC were improved by the addition of phenolic resin confirming the reinforcing effect of this filler. On the contrary, the impact strength diminishes with increasing amounts of phenolic resin, although this property is strongly dependent on the particle diameter.

SEM micrograph of the nylon 6‐polyesteramide block copolymer (80/20).