【作者】 李军；

【导师】 喻胜华；

【作者基本信息】 湖南大学 ， 数量经济学， 2004， 硕士

【摘要】 随着全球经济的迅猛发展，知识的占有逐渐成为市场经济主体成功角逐市场的关键，然而这种知识绝然不同于平常意义上的知识，它必须具有开创性和可靠性两个基本特征，开创性是指知识是新发现的，可靠性是指知识能积极辅助判断和决策的品质。 针对知识的获取，本文利用上市公司披露信息数据库为平台，系统地研究了数据挖掘中的关联规则挖掘和神经网络方法，建设性地运用数据挖掘方法去获取基于上市公司的知识，克服了国内以往数据挖掘研究“极大忠诚于外国文献”、重理论轻实践的缺点，生动、形象地展现了基于关联规则挖掘和神经网络方法的知识发现过程。 在实证研究中，通过挖掘出的有趣规则——“上市公司被处罚→上市公司被特别处理”，我们获得知识——“如果某上市公司被处罚，那么该上市公司未来有80％的可能性会出现财务危机”。这为投资者和监管部门的决策提供了有用信息。通过对把神经网络方法应用于财务风险识别的研究，我们不仅把模型的仿真度提高到100％，而且显著提高了财务状况特征识别准确率。这将为神经网络模型运用于社会科学领域提供了切实可行的经验。 理论上而言，研究不仅把数据挖掘和上市公司的知识发现联系在了一起，而且，前所未有地，就数据挖掘在上市公司知识发现中的运用做了比较系统的阐述，抛砖引玉地把上市公司相关信息的数据挖掘工作呈现在人们面前，开拓出了一个广阔的实证分析领域。更多还原

【Abstract】 With rapid development of global economy, occupying knowledge gradually become the key to succeed in market competetion for market economy subjects. Otherwise, this kind of knowledge is significantly different from the ususl knowledge by two features of it-"initiativeness and reliability".The initiativeness means that it is the first time that it is discoveried,and the reliability mean to its ability to help positively people make discriminant and decision.For that, this thesis, based on the database that contain the information issued by listed A share companies, not only systematically make research of Association Rules Mining and Neural Network Prediction Method in Data Mining,but also acquire some knowledge about listed A shares companies by constructive use of these two Datamining tools. Overcoming the drawback of "greatly loyal in foreign documents"and empirical analysis lackness tagged to Datamining research in China,this paper vividly represent Knowledge Discovery process through empirical research of Association Rules Mining and Neural Network Method.The finding through Association Rules Mining shows knowledge that the probability is 80% of that one listed company will present financial crisis in the future if this listed company has been forfeited. Obviously ,it transmit some valuable information to not only investors but also the government supervisors. Not merely this, the empirical study on Neural Network indicate that the company financial situation detection model based on Neural Network not only bring the fidelity up to 100%,but also significantly highten the accuracy ratio of detection. It is no doubt that the finding offer feasible experience to help people apply the Neural Network into detecting objective subject in social science field.In theory, the thesis cast the problem How to discovery in Database of listed companies to attract more and more people study on it ,then open up a wide empirical analysis field, while firstly make systematical research on Datamining application into Knowledge Discovery in listed companies.更多还原