Application of support vector machines regression in prediction Shanghai stock composite index

0　IntroductionNeuralnetworks,inrecentyears,havebeensuccessfullyusedformodelingfinancialtimeseries.Researchesrevealneuralnetworkshavebetterresultsindescribingdynamiccharac teristicscomparingwithclassicalstatisticalmethods[1 3] .Back propagation (BP)neuralnetworkisamostpopularlyusedneuralnetworkmodelintheareaoffinancialforecasting .AlthoughBP’sstructureissimple ,ithaspowerfulabilityinsolvingvariousproblems.TherearealsoshortcomingswithBP :needsformanycontrolparameters,difficultyingettingsta…     

Specific interactions of the adenovirus proteinase with the viral DNA,an 11-amino-acid viral peptide,and the cellular protein actin

The adenovirus proteinase (AVP) is synthesized in an inactive form that requires cofactors for activation. The interaction of AVP with two viral cofactors and with a cellular cofactor, actin, is characterized by quantitative analyses. The results are consistent with a specific model for the regulation of AVP. Late in adenovirus infection, inside nascent virions, AVP becomes partially activated by binding to the viral DNA, allowing it to cleave out an 11-amino-acid viral peptide, pVIc, that binds to AVP and fully activates it. Then, about 70 AVP-pVIc complexes move along the viral DNA, via one-dimensional diffusion, cleaving virion precursor proteins 3200 times to render a virus particle infectious. Late in adenovirus infection, in the cytoplasm, the cytoskeleton is destroyed. The amino acid sequence of the C terminus of actin is homologous to that of pVIc, and actin, like pVIc, can act as a cofactor for AVP in the cleavage of cytokeratin 18 and of actin itself. Thus, AVP may also play a role in cell lysis.Received 14 November 2002; received after revision 28 April 2003; accepted 30 April 2003     

基于改进BP算法的刚框架连接损伤识别方法

运用权值、阈值同步学习BP算法(简称为改进BP算法)对钢框架连接损伤进行识别,并且选用由结构正则化的频率变化率和少数点的振型模态分量构成的组合参数作为网络输入参数.通过对一个4层钢框架数值模拟,比较了改进算法和传统算法测试结果,从而证明改进算法更加精确、优越.     

BP网络与NRTL方程在汽液平衡数据预测中的比较研究

用BP神经元网络,对乙醇-环己烷二元体系的汽液相平衡数据进行关联和预测,并用三参数NRTL方程的关联及预测结果进行对比研究.结果表明,BP神经元网络对汽液平衡数据的关联和预测结果在精度上均明显优于NRTL方程。     

真空助力器带主缸总成疲劳试验系统辨识

真空助力器带主缸总成是汽车制动系统的关键增力部件,其气动疲劳试验系统由气动子系统、液压子系统、电控子系统等多个子系统构成,属于复杂的非线性系统。该系统存在众多不确定因素,因此无法直接建立数学模型。为了考察该系统的动态性能,采用递归BP神经网络对该系统进行辨识。试验结果表明,递归BP神经网络能够很好的逼近该复杂非线性系统,在不同时间序列作用下均有较好的泛化能力。     

铝型材挤压模具导流孔结构优化

基于MATLAB平台,将BP人工神经网络、遗传算法和数值模拟技术应用于铝型材挤压模具的导流孔形状优化设计。由正交实验法安排模拟实验组合,采用SuperForge软件对进行型材挤压过程进行数值模拟,并以挤压时金属流出模口平面的z向质点流速的均方差作为模型目标值;将模拟结果作为人工神经网络的输入样本对进行网络训练并建立网络知识源;通过遗传算法求得模型的全局优化解;最后通过有限体积法数值模拟技术验证并比较优化所得导流孔形状与经验法确定的导流孔形状对金属流动均匀性的影响。分析结果表明,通过调整导流孔形状能使金属流出模口的速度分布更均匀,表明对挤压模具导流孔形状的优化是有效的。     

Recent temperature increase recorded in an ice core in the source region of Yangtze River

Interests on climate change in the source region of Yangtze River have been raised since it is a region with the greatest warming over the Tibetan Plateau (TP). A 70-year history of precipitation δ~(18)O has been recovered using an ice core record retrieved in a plat portion of the firn area in the Guoqu Glacier (33°34′37.8″N, 91°10′35.3″E, 5720 m a.s.l.), Mt. Geladaindong (the source region of Yangtze River), in November, 2005. By using a significant positive relationship between ice core δ~(18)O record and summer air temperature (July to September) from the nearby meteorological stations, a history of summer air temperature has been reconstructed for the last 70 years. Summer temperature was relatively low in 1940s and high in 1950s to the middle of 1960s. The lowest temperature occurred in the middle of 1970s. Temperature was low in 1980s and dramatically increased since 1990s, keeping the trend to the begin-ning of the 21st century. The warming rate recorded in the ice core with 0.5℃/10 a since 1970s is much higher that that in the central TP and the Northern Hemisphere (NH), and it becomes 1.1℃/10 a since 1990s which is also higher than these from the central TP and the NH, reflecting an accelerated warm-ing and a more sensitive response to global warming in the high elevation region.     

BP神经网络算法在龙门吊称重系统的应用

介绍了龙门吊称重系统的应用,运用BP神经网络的学习算法对称重系统进行校正,与传统加权平均算法进行了比较,其结果对精度具有较好的提高。     

基于BP神经网络对念珠藻摇瓶培养条件的优化

利用均匀设计就pH、装液量、接种量和光强等影响念珠藻生长的因素进行研究,并以所得的数据为样本建立一个结构为4-6-1的神经网络.当网络的学习速率为0.08,动态参数为0.8,SIGMOID参数为0.8,隐含层神经原数为6时,网络的拟合平均误差为2.24%.在固定装液量为100 mL的条件下,利用该网络优化念珠藻的培养条件,得到最优的培养条件为pH7.4,接种量1.5 g/L,光强5 700 lx,生物量达到1.702 3 g/L.     

基于神经网络的简支梁桥预应力衰减评估模型

以后张法室内模型梁受弯全过程试验采集的数据为样本,研究了预应力混凝土桥梁有效预应力与各主要影响因素之间复杂的非线性关系,应用改进BP神经网络建立了混凝土桥梁服役期内预应力衰减的神经网络评估模型。该模型以受弯梁全过程P-Δ理论曲线为依据,构造了基频衰减率(IFDETR)、变形指标(IDISP)、混凝土应变变化率(ICSV)、钢筋应变变化率(IRSV)4种归一化指标作为网络输入值,以计入时变效应的预应力衰减率指标(IEPDR)作为网络输出值。测试样本的评估结果表明,误差小于1%的测试样本数在样本集中所占的比例大于85%,且误差均小于单因素回归方法。