遗传算法优化BP神经网络的制造业上市公司财务预警研究

发布时间：2018-06-06 00:00

本文选题：财务预警 + BP神经网络　；参考：《河北大学》2014年硕士论文

【摘要】：随着我国市场经济发展程度的提高，市场竞争愈发激烈，企业遇到了更多危机与挑战。对于为数众多的制造业上市公司而言，财务问题尤为重要，一旦财务陷入困境，不但危机自身的生存与发展，也给投资者、债权人以及众多利益相关者带来重大损失。因此，对于财务状况进行预警研究与探索，具有重要意义。国内外的大量文献已经能够证明，企业的财务指标对企业财务发展状况有一定的预示作用，但是对于财务预警指标体系的构建是不完善的。在研究财务预警的大量模型中，，也存在许多不足和局限性。本文就是针对财务预警指标体系的构建和预警模型来进行研究和探索。通过分析前人的研究成果，总结出了28个对企业财务状况影响较大的预警指标。其中，包含了财务指标和公司治理等非财务指标。丰富了预警指标体系，为财务预警模型研究提供了基础。当输入变量过多，输入变量之间不是相互独立时，BP神经网络预警模型容易出现过拟合现象，从而导致模型精度过低、建模时间长等问题，因此本文选用遗传算法对已经选取的预警指标进行变量降维，并对BP神经网络预警模型进行优化。实证分析分为两步，第一步先用BP神经网络做预警分析，第二步应用遗传算法优化BP神经网络做预测。通过对比两步的计算结果，发现用遗传算法优化过的BP神经网络对财务进行预警有更明显的效果。大大提高了预测精度，缩短了建模时间，为财务预警的理论和实践研究提出了新的思路。
[Abstract]:With the development of the market economy in China, the market competition is becoming more and more intense, and the enterprises have encountered more crises and challenges. For a large number of manufacturing listed companies, financial problems are particularly important. Once the financial crisis is in trouble, not only the survival and development of the crisis itself, but also the investors, creditors and many stakeholders are also brought. Therefore, it is of great significance to conduct early-warning research and Exploration on the financial situation.
A large number of documents at home and abroad have proved that the financial indicators of enterprises have a certain predictive effect on the financial development of enterprises, but the construction of the financial early-warning index system is not perfect. There are also many shortcomings and limitations in the study of a large number of financial early-warning models. This paper is aimed at the structure of the financial early warning index system. Construction and early warning model for research and exploration.
Through the analysis of previous research results, 28 early warning indicators have been summarized, including financial indicators and corporate governance, which enrich the early warning index system, which provides a basis for the study of financial early warning models. When the input variables are too much, the input variables are not independent, BP The neural network early warning model is easy to appear over fitting phenomenon, which leads to the low precision of the model and the long modeling time. Therefore, this paper uses genetic algorithm to reduce the variable of the selected early warning index, and optimizes the BP neural network early warning model.
The empirical analysis is divided into two steps. First, the BP neural network is used to make early warning analysis, and the second step is to optimize the BP neural network by genetic algorithm. By comparing the results of the two steps, it is found that the BP neural network optimized by genetic algorithm has a more obvious effect on the financial early warning. It greatly improves the prediction accuracy and shortens the modeling time. A new train of thought is put forward for the theory and practice of financial early-warning.
【学位授予单位】：河北大学
【学位级别】：硕士
【学位授予年份】：2014
【分类号】：TP183;F406.72

【参考文献】