定性映射及定性转化程度函数在财务分析中的应用
详细信息    本馆镜像全文|  推荐本文 |  |   获取CNKI官网全文
摘要
企业综合实力是指企业在较长时期內的市场竞争能力,不仅包括现在的生存状态,也包括将来的发展前景。作为理性的投资者,正确认识和评价企业披露的信息是有必要的。目前我国上市公司会计年报中包括了一些财务指标和经营业绩指标,对考察一个企业的经营现状提供了有用的信息,但不足以准确全面地分析其经营成果、机制创新和发展潜力。
     本项目试图建立更为一般性的企业综合实力评价体系,满足投资者和管理者的需要。通过定性和定量的综合的分析,可以区分企业的综合实力等级及其在同行业中所处的位置。本项目对股票分析也提出了一个解决方法。股票信息瞬息万变,从大量的股票数据中找到真正有用的信息是金融领域的一个难题。目前市面上主要通过各种参数指标的组合去描绘股市变化的规律,然而,现实情况是,人们用自己熟悉的某些参数,如:MACD、KDJ等进行的指标分析,大多都不准确。要么提前、要么滞后。至于某些人煞费苦心试图寻找的所谓“万能指标”,至今仍不知在何处。
     一般认为,股票价格及其变化趋势反映的是上市公司的综合实力,怎样根据股票价格及其变化趋势,去理智地分析、评估一家上市公司的综合实力,并对该公司的“好”或“坏”作出一个准确的判断,对投资能否赢利来说是很重要的。
     因为综合实力是反映上市公司“好”或“坏”的重要属性,股票价格及其变化趋势是反映综合实力(这一属性)的量特征,而“强”或“弱”则是反映综合实力(这一属性)本质的两个质特征,因此,从属性论的观点看,股票分析和决策的整个过程,就是一个事物属性的量——质特征转化过程。
     基于这样一种认识,本文利用属性论方法提出的的定性映射及其转化程度函数(Conversion Degree Function,mDWCDF)η_i(x,α_i,β_i,ξ_i)既能解决属性量—转化差异,又能确定不同属性间的权重的特点,设计了一个财务分析系统,并给出了层次分析—微变因子调控算法和均分斜率趋势分析算法两个算法,给出了财务分析中的主指标权重确定和主指标预测的一种新的分析方法。
    
    定性映射及其转化程度函数在财务分析中的应用
     1.层次分析一微变因子调控算法
     设嘿为专家关于财务主指标的基权值的向量空间,Z二(t,,八,t。)。衅为使用层
    次分析法(AHP法)得到的第i位专家关于财务主指标的基权值向量,其次,对向量
    空间嘿中的向量进行统计平均,得到统计平均后的基权值向量E=(e,,A,e,)o然后
    取调控因子向量入沮一(△lrl,AA,Am气)其中八E(一1,1),将向量E与向量入沂
    相加,使得权值向量变为E一协+八rl,AA,氏+么八),最后通过数据调整向量△贝
    的值,当向量△趋于无穷小时就得到一个包含k(k>l)条向量的切线丛,则该切线丛
    内的向量可认为是主指标权值的较优解。
     2.均分斜率趋势分析法
     首先假设财务主指标的走势曲线为:P(x)一bl十bzx+·+气广一‘(脚    采用多项式数据拟合,求出:b‘,i二l,A,m。其次,将拟合方程根据属性论方法中的
    模式—向量转换特性转换为向量。然后构造一个人工神经元网络,然后根据定性基
    准的w_内积变换与人工神经元的关系由该人工神经元网络得到一个定性基准的w_内
    积变换,将向量输入进行学习,得到训练范例,然后将己有的向量输入到该定性基准
    的w_内积变换中,得到匹配的向量。我们由该匹配的向量可得到预测点t+1点处的
    值与拟合点t点处的值的大小关系。最后利用差分方程来求得预测点t+l点处的值。
    因此可对当前财务主指标曲线的走势进行预测。
The company comprehensive competence which is the market competent ability in a long-term period includes not only survival situation now but also the future aspect. As a rational investor, recognizing and assessing successfully the information of the company is essential. In current the accounting reporters of the companies in our nation include some financial indexes and achievement indexes which provide useful information to assess the situation of the company but not enough information to analysis comprehensive achievements and innovation and latent capacity.
    Our project tries to establish an ordinary assessing system about company comprehensive competence to satisfy the need of investors and administrators. Through comprehensive analysis of attributive from quantity to quality, we can distinguish the level the company in. And our project provides a solution of certificate of stock analysis. The information of certificate of stock is so various that it is difficult to find the real useful information even to the economists. The main approach currently depicting the currentness of the certificate of the stock is the association of the parameters. But many people have found that to analysis with the facility of associating-parameters of MACD, KDJ , etc. is inaccurate which may be advanced or lagged. And it is not unrealistical to search a universal index sign.
    Because the stock is the reflection of the integrated competence,it is important for a successful investment how to analysis and assess a company's
    
    
    comprehensive competence according to the stock price and the change currency and then give a accurate determination about the company.
    Because the comprehensive competence is the major index depicting the qulity of a company and the price and change currency of stock is the quantity character of the comprehensive ability, according to which the "good" or "bad" is the quality character of the comprehensive competence, the stock analysis and decision is a quantity-quality changing process.
    Our project use The Qualitative Mapping model and which induced m_Dimension Weight Conversion Degree Function which could resolve the conversion difference from quantity to quality of attributive in the Attribute Theory and could certify the weights of different attributes to design a financial system and give two algorithms: one is Layerevel Analysis-Gradualchange Factor Controlling and the other is Equipartition Slope Controlling. So we can use these methods to give a new analysis way to certify the main weight in a financial analysis and to predict the main weight.
    the algorithm of Layerevel Analysis-Gradualchange Factor Controlling
    Firstly, we hypothesise that Mnm is the vectorial space of fundamental weight about the financial main index signs, and T: = (t} ,A ,tm)M"m is No. i fundamental weight vector which we get through the method of Layerevel Analysis (AHP).
    Secondly, we make a statistical average of all the vectors in the Mm ,and
    hence get the statistic fundamental vector 7 = (e} ,A ,em).
    Lastly, we take a Controlling factor's vector { =(Ai,'A Ar}' A e 1, 1) .We add the vector E and the vector ,then make vector turn into =
    Ari'AA '6m+Amrm)-We can injustify the value of AR ' wnen vector is
    diminished, and get a clump of tangents which number is k(k>1). We can say that above tangents are better solutions.
    
    the algorithm of Equipartition Slope Controlling
    Firstly, we assume that the curvature depicting fianancial currency is:
    we adopt the method of
    polynome-data-fitting and hence obtain the consequence which is bj,j=1,...,m .
    Secondly,we turn the fitting equation into vector according to the model-vector conveying character in the The Qualitative Mapping.And we construct a manul nerve net.then we can get a m_Dimension Weight Conversion Degree Function from the net according to the relations between the m_Dimension Weight Conversion Degree Function and the manul nerve net. We input the vector and obtain the train model, finaly we obtain the fitting vector. So we can use the fitt
引文
1 叶中行 陆培丽 基于因子分析和和自组织映射的公司财务分析 第十三届中国神经网络学术会议论文集 人民邮电出版社
    2 林建忠、叶中行,《数理金融》,科学出版社出版 (1998.12)。
    3 叶中行,顾立庭,“股市变化模式的两种神经网络方法”,上海交通大学学报。”Vol.29,No.2, P100-104,(1995).
    4 林建忠,叶中行,“推广线性二阶抛物型方程Cauchy问题的Feynmam-Kac定理”,上海交通大学学报,Vol.34,No.4,582-588 (2000.4)。
    5 洪楠 侯军 SAS for Windows统计分析系统教程 电子工业出版社
    6 李东风 SAS多元统计分析 电子工业出版社 1997:130~162
    7 廖泉文 人力资源考评系统 电子工业出版社 1997:25~34
    8 冯嘉礼,董占球,基于属性整合的知觉模式生成与识别模型,计算机研究与发展,1997,34(7):487-491
    9 姚伯茂,质量互变规律,质与量[M],中国大百科全书,哲学卷Ⅱ,中国大百科全书出版社,1987:1180~1181
    10 冯嘉礼,定性映射诱导的模糊人工神经元和网络[J],南京大学学报,2003,39(2):172~181
    11 Jiali Feng, Degree Functions and Fuzzy Artificial Neurons Induced By Qualitative Yapping[C], Proceedings of International Conference on Fuzzy Information Processing Theory and Application, FIP03' s, Tsinghua University Press & Springer, March'511-517
    12 潘国美 预测方法 电子工业出版社 1997:35~50
    13 冯嘉礼,赵唐和黄伟杰,基于属性量-质特征性转化及其定性映射的KDD型,计算机研究与发展,2000,Vol.37(9):1114-1119
    14 冯嘉礼,判断基准的可变性与面向判断的性质逻辑.广西师范大学学报,1995.1(12)1-5
    15 冯嘉礼,冯嘉仁,詹增修.以属性为基础的知识库建库原则.计算机研究与发展,1987.11(24):55-61
    16 冯嘉礼.心理认知结构与性质多面体重心坐标系.广西师范大学学报,1989.2:1-6
    17 冯嘉礼.一种会学习的感知机决策机模型.广西师范大学学报,1992.1(10):1-6
    
    
    18 冯嘉礼,叶中行,刘永昌.神经检测映射、感知同态及范畴同态引理,国防科技大学学报,1995,Vol.17 Sup.:4-7
    19 冯嘉礼.基于性质坐标系的一种非单调推理与决策.大自然探索,1990.4:87-93
    20 潘谦红,王炬,史忠植.基于属性论的文本相似度计算.计算机学报,1999.6(22):651-655
    21 Grabisch M. Fuzzy intergral in multi-creria decision making. Fuzzysets and Systems, 1996(66):279-298
    22 Foor J C, Roubens M. Aggreation and scoring procedures in multicreteria decision making methods. Pro. Lst FUZZIEEECong. San Diego, CA, 1992,1261-1267
    23 French, S. Multi-Attribute Decision Analysis after a Nuclear Accident
    24 李洪兴.因素空间理论与知识表示的数学框架(Ⅳ).系统工程与数学,1995.1(14)
    25 李洪兴.因素空间理论与知识表示的数学框架(Ⅷ).模型系统与数学,1995.3(9):1-9
    26 汪培庄.模糊集与随机集落影.北京:北京师范大学出版社,1985.9
    27 梅绍祖.模糊控制与变权的确定.系统工程理论与实践,1995.5
    28 李洪兴.因素空间理论与知识表示的数学框架(Ⅶ).模糊系统与数学,1995.2(9):16-24
    29 Bryson N. Mobolurin A. An action learning evalution procedure for multiple criteria dicision problems. European Journal of Operational Research, 1996.96:379-386
    30 冯圣红.一种多指标综合评价合成技术方法研究.模糊系统与数学,1999.2(13)
    31 冯嘉礼,詹明,叶中行.感知和判断中的基准变换及其性质坐标分析法.广西科学,1994.1(4):6-13
    32 冯嘉礼.内积的特征向量与常数的内积分解.广西师范大学学报(自然科学版).2001.19(1)
    33 冯嘉礼.从属性量检测到质特征定性的感觉特征抽取模型.广西师范大学学报(自然科学版).2001.19(2)
    34 赵光武,思维科学研究,中国人民大学出版社,1999:13
    35 钟义信,信息科学原理,北京邮电大学出版社,1996,24
    36 Semir. Zeki, The Visual Image in Mind and Brain, Scientific American, Sept 1992, Vol. 267(3): 69-76
    37 冯嘉礼,詹蒙,感觉数据_特性抽取的定性映射模型,清华大学学报,1998,S2:19-23
    38 冯嘉礼,詹蒙,感觉诸映射间的关系结构,计算机工程与科学,1999,21(2):1-6
    39 冯嘉礼,以属性为基础的知识库建库原则,计算机研究与发展,1987,24(11):56-61
    
    
    40 董占球、冯嘉礼,按模式记忆理论的数学描述(Ⅰ)—记忆模式的属性坐标表示法,计算机研究与发展,1998,35(8):694-698
    41 冯嘉礼等,感知与判断中的基准变换及其性质坐标分析法,广西科学,1994,4(1),p6-13,22

© 2004-2018 中国地质图书馆版权所有 京ICP备05064691号 京公网安备11010802017129号

地址:北京市海淀区学院路29号 邮编:100083

电话:办公室:(+86 10)66554848;文献借阅、咨询服务、科技查新:66554700