金融时间序列分析与建模

介绍了金融时间序列分析与建模的主要理论、方法,着重论述了近20多年发展起来的非平稳时间序列的建模方法及工具,指出进一步的研究方向.最后,对上证综合指数这一金融时间序列进行了实证分析.     

Pregelation Behaviour of Coagulation Processes with the Constant-Reaction-Number Kernel

We propose an irreversible binary coagulation model with a constant-reaction-number kernel, in which, among all the possible binary coagulation reactions, only p reactions are permitted to take place at every time. By means of the generalized rate equation we investigate the kinetic behaviour of the system with the reaction rate kernel K（i;j） = （ij）^w （0 ≤w〈1/2）, at which an i-mer and a j-mer coagulate together to form a large one. It is found that for such a system there always exists a gelation transition at a finte time to, which is in contrast to the ordinary binary coagulation with the same rate kernel. Moreover, the pre-gelation behaviour of the cluster size distribution near the gelation point falls in a scaling regime and the typical cluster size grows as （to - t）-1/（1-2w）. On the other hand, our model can also provide some predictions for the evolution of the cluster distribution in multicomponent complex networks.     

Double-Humped Transverse Density Profile in Two-Dimensional Chute Flow with Rough Sidewalls

We study a two-dimensional granular rapid flow with rough sidewalls stuck with the same size discs by molecular dynamics simulation. A transient state of the double-humped density profile in the flowing process has been found, which appears and moves as travelling wave and is the same as the phenomena in the recent experiments [Acta Phys. Sin. 53 （2004） 3389 （in Chinese）]. Our simulation shows that the rough sidewalls play an important role in the converting momentum of boundary discs from the vertical direction to the horizontal one through particle collisions to form this profile and the good elasticity of discs ensures this effect. The appearance of the double-humped profile may be a precursor, which determines if the whole flow will be far repulsed from the boundary and become dilute eventually.     

g-稳定性与 R~n-稳定性的等价性

一、引言近年来,相空间理论得到了深入的发展.人们找到了一些便于应用的具体的相空间如 C_g,C_h 等,其中 C_g 是研究较多的一个空间.研究表明 C_g 具有优良的性质,但 C_g 如不附加进一步条件的话,并不是 Hale-Kato 意义下的允许空间,故[1,2]的稳定性的等价性定理并不能直接应用于 C_g.鉴于 C_g 在理论和应用上的明显的重要性,本文研究了在 C_g 和 R~n 中泛函微分方程解的稳定性的等价性问题.研究表明 C_g 在这方面也具有良好的特性.作为应用,本文还研究了 Volterra 积分微分方程周期解的存在唯一性及稳定性问题,得到了新的有趣的结果.     

一种区间数的因子分析技术及其在证券市场中的应用

传统的因子分析技术能够有效地对高维变量空间进行降维处理,但它对于样本空间却缺乏行之有效的降维效果.为了解决这一问题,一种针对大量样本数据、新的因子分析技术———区间数因子分析技术(intervaldatafactoranalysis,IFA)被提出并得到了迅速的发展。IFA方法对传统的数据概念做了本质性的扩张,运用'数据打包'的理念,对海量原始数据在不破坏其原有内在逻辑关系的前提下,可以进行变量和样本点维度的双重降维。本文详细阐述了区间数因子分析技术的原理,并以中国股票市场为案例研究背景,结果表明IFA分析技术对大规模多维数据系统做综合简化是十分有效的。     

The Spaces b~p with 0 < p < 1

Let 0 < p < 1. In this paper we prove that the Bergman norm on the p-th Bergman space bp(H) is equivalent to certain "normal direction norm" as well as certain "tangential direction norm", where H = Rn-1 × R is the upper half space. As an application, we get the boundedness of harmonic conjugation operators on bp(H).     

四阶累积量的MUSIC算法在海流测向中的应用

针对高频地波雷达系统OSMAR2000海流测向中海流元数不易估计的问题，利用四阶谱累积量MUSIC算法的空间谱结构特点，采用试探法确定信号源数并估计信号源的位置．通过对模拟海流和实测数据的处理实验表明，利用该算法可以较准确地估计出海流元的个数和方位，并且比传统MUSIC方法获得更小的角度估计方差，有效地反演了海洋表面流信息．     

BP人工神经网络用于除草剂活性预测

采用AMI方法研究了10种新型的磺酰脲类除草剂的电子结构，并以原子的Mulliken净电荷和除草剂在不同浓度(100，10mg／L)下对油菜、稗草两种作物的根、茎部位的抑制率为训练样本集。构造并训练得到具有活性预测能力的BP神经网络．结果表明，该BP网络不仅能对训练样本很好拟合。亦能对未知化合物的活性作出很好的预测．     

O～3岁儿童微量元素检测及纠正150例观察与体会

对西安地区150例0～3岁儿童进行了六种微量元素：钙、锌、铜、铁、镁、硒的检测，发现微量元素缺乏可导致急慢性感染、免疫力低下、小儿营养不良、贫血、发育迟缓等症状。并对小儿营养不良、喂养不当进行矫正。其中铁、钙、锌同时缺乏占50％。