基于加权特征子空间的支持向量机核函数研究

针对数据分类交叉空间易错分问题,提出一种基于加权特征子空间的支持向量机核函数方法。该方法利用加权特征子空间与稀疏表达等相关理论进行支持向量机核函数优化,首先利特征子空间重叠率与数据的信息熵对数据特征进行加权,再通过对L_1范数正则项的加权处理调节异类数据间的稀疏性和同类数据间的稠密性;最后对处理好的数据进行分类测试。仿真实验表明该算法能够在一定程度上提升分类效果,以达到优化核函数的目的。     

我国金融服务业的软实力影响因素分析

随着改革开放的不断深化,我国金融业快速发展,金融实力显著提升。通过阐述"软实力"的概念,探讨我国金融软实力影响因素。从国家软实力、文化软实力、城市软实力出发,研究软实力的真正内涵,结合金融业特征,运用决策试行与评价实验室分析法,对金融软实力的影响因素进行了深入的分析,认为金融政策、金融规模、金融人才等对金融软实力具有很大影响,并针对我国提升金融软实力提出了建议。     

现代农村物流金融中心诸利益主体的博弈分析

现代农村物流金融的特殊性,要求现代农村物流金融中心的建设必须由地方政府、物流企业、商业银行、民间资本或境外资本持有者等诸利益主体共同参与来完成,诸利益主体的行为不同程度地影响着现代农村物流金融中心的建设与完善。本文运用多人合作对策理论,建立一个四方博弈模型,分析现代农村物流金融中心建设过程中诸利益主体间应如何分配其合作收益,这无疑对现代农村物流金融中心的建设,发展现代农村物流,发展新农村经济有着极其重要的意义。     

F#-Time, Functions and Processes: Towards Sustainable Investment Risk Benchmarks

‘I made a mistake’: Alan Greenspan (Financial Times: Alan Beattie and James Politi: Washington, 23rd October 2008). Such are the words of great men, for even in troubled times their self-effacing manner provides useful guidance. Whilst Mr Greenspan may feel this way, he is a product of his environment, one that has seen the cumulative development of financial instruments and strategies that have not been thought through as to their impact on a complex economy. Mainly this is because risk is thought to be discrete and the methods used to price it are flawed. To an engineer the control of a machine is built-in. Although the economy is not a machine, but an intensely connected complex of ever emerging businesses, the process of control needs to be structured in a similar manner. Pricing investment risk in this environment should never have been left to opaque institutions, or processes that do not recognise the co-dependencies of business and systemic functionality. To do so is to ignore the correlation of events in a highly connected world. These events are dynamic and conditional, whose outturns are unknowable. This does not mean unmanageable, but that the control process be built-in to businesses and government in a consistent manner, transparent yet using different parameters. Transparent means that data, assumptions and processes need to be monitored and published in timely manner. As far as accounting for results is concerned it should be recognised that budgeting and reporting to investors is founded on dynamic processes that are therefore changeable; usually out of date; and co-dependent upon others within a complex dynamic network that is both internal and external to the business. The works of Stafford Beer (Brain of the Firm, Heart of the Enterprise, Diagnosing the System) Fredrick Vestor (The Art of Interconnected Thinking) and others are examples of how to manage the internal dynamics of a business and point to a methodology that synthesises the approaches of investors such as Warren Buffett so that extreme outcomes such as the Credit-Crunch 2008 can be reduced in frequency but investors are free to ‘take their risks’. This research aims to compare two extreme events in the financial arena, the ‘Reinsurance Spiral of the late 1980s’ and the ‘2008 Credit-Crunch’, show their commonalities and propose methods that supply liquidity in all but gross systemic failure and allow investment risk to be more ably assessed and priced. It is not meant to be an exhaustive analysis but one focused on how ignoring the proper relationship of time, functions and processes brought about the current problem in both insurance and the capital markets and how a solution may be found. This research note offers an overview on the ongoing PhD research on the topic.

软测量技术进展

软测量技术是二十世纪六、七十年代涌现的一种为控制过程建立模型的新技术,被列为未来控制领域需要研究的几大方向之一。本文主要从软测量多模型结构和支持向量机在软测量中的应用等方面对近几年来软测量技术的进展进行了介绍,并提出进一步的思考。     

我国金融消费者法律保护的完善

世界各国在后金融危机时代,已经掀起对金融消费者的法律保护高潮。我国当前对金融消费者的法律保护还处在萌芽状态。因此,在梳理我国金融消费者法律保护的现状下,借鉴域外的相关法律保护制度,完善我国的金融消费者权益具有一定的理论和实际意义。     

遗传算法选择性集成多分类器的企业财务困境预测

为了更加有效地进行企业财务困境预测,本文提出了基于遗传算法选择性集成的多分类器系统。与事先静态给定系统内部基本分类器组成不同,该方法以组合系统预测准确率为优化目标,无需度量各基本分类器之间的差异性,可以动态挖掘最优组合系统。实证研究中以中国上市公司为研究对象,以10折交叉验证准确率作为评价标准,结果表明该方法显著优于全集成以及单分类器最优模型。     

基于最近邻法和支持向量机的个人信用评估方法

本文针对两类个人信用数据混叠较严重的数据集,提出对数据集先利用最近邻算法进行修剪,再应用SVM算法对个人信用进行评估的NN—SVM方法。仿真实验表明基于NN—SVM算法的个人信用评估方法比直接用SVM算法进行分析来的更加准确。同时,对比RBF_LS—SVM,Linear LS-SVM,Region single tree等算法的结果,发现NN—SVM算法明显优于其它算法。     

简析中古英国财政机构的演变

财政机构的演变是透视中古英国王权性质变化的重要视角。中古英国财政机构经历了一个由王室财政机构向国家财政机构的演变过程,它揭示了带有较强私权性质的王权向不断扩大的公权性质的王权的转变过程。     

基于振动特征估计的气体绝缘开关设备故障检测与定位

为了有效检测气体绝缘金属全封闭开关设备(Gas Insulated Switchgear, GIS)故障,提出一种基于振动特征分析的GIS设备故障检测与定位方法。首先,选取GIS金属外壳多测点振动信号作为设备的状态变量。然后,构建多输出支持向量回归(Multi-Output Support Vector Regression, MOSVR)模型对GIS设备多测点振动特征进行回归估计,并利用多测点振动特征残差相对值计算故障预警指标。之后,利用指数移动加权平均方法计算自适应阈值来衡量故障预警指标的变化趋势以实现设备的故障检测。检测出故障后,利用箱形图分析多测点振动特征残差来定位设备故障。最后,利用现场实测信号验证了提出方法的有效性。