基于改进的BP神经网络的农户小额信用贷款风险评估模型研究
发布时间：2018-11-01 17:40
【摘要】：随着农村经济的飞快发展,农村的信用贷款需求也跟着增多。农村的信用贷款以农户小额信用贷款为主。而农户信用风险评估在农户小额贷款中至关重要,它的好坏制约着农村经济的发展。因此对于农户小额贷款风险评估系统的研究十分必要。 本文首先介绍了农户信用评价的概念,以及农户小额信用贷款的概念,接着详细分析了国内外对农户小额信用贷款风险评估的研究,通过分析发现国内在这方面的研究不如国外。由于人工智能在信用评价应用中有很大优势,在国外已经广泛使用人工智能技术。而BP神经网络技术是人工智能技术中的一个重要技术,它有着较强的学习和自适应能力、较好的内在并行计算和存储,是一种稳定的非线性方法等优点,所以BP神经网络在农户小额信用贷款风险评估的研究中也得到应用。然而传统的BP神经网络有着收敛速度慢、易陷入局部极小等不足。针对这些不足,人们先后提出了很多改进的策略,比如加动量项、自适应学习速率、LM算法、人工免疫、遗传算法、粒子群优化算法等。 本文使用了一种新兴的群体智能算法量子粒子群优化算法(QPSO)改进BP神经网络模型。量子粒子群优化算法(QPSO)有着调节参数少、简单易实现的优点,并且有着较好的收敛性能和全局搜索能力,能在一定程度上能够克服BP神经网络算法在收敛性能上的不足。 通过加入自适应变异对量子粒子群优化算法(QPSO)进行改进,并取得了很好的效果。由于量子粒子群算法在早期收敛速度较快,所以在后期可能没有达到全局最优时已经聚集到某一点,形成局部极小。针对这一缺点加入自适应变异对量子粒子群优化算法(QPSO)进行改进。 本文改进后的量子粒子群优化算法(QPSO)通过优化BP神经网络的权值和阈值,从而得到改进的BP神经网络模型。它与用遗传算法等改进的BP神经网络模型相比,能够更有效的提高BP神经网络的收敛速度,防止陷入局部极小。 然后,将改进后的BP神经网络模型应用到农户小额信用贷款风险评估系统实验中。在仿真模拟实验中,从数据中随机抽取5组数据集进行实验,然后再取这5组实验结果的平均值与用传统BP神经网络进行对比,我们可以很明显的看出,经过改进的BP神经网络更能帮助我们提升信用评估过程中的精确性,错误的可能性也可以降低。 本文最后从农户小额信用贷款风险评估的研究方面和改进BP神经网络方面进行了下一步展望。
[Abstract]:With the rapid development of rural economy, rural credit loan demand also increases. Rural credit loans to farmers mainly small credit loans. The credit risk assessment of farmers is very important in farmers' small loans, which restricts the development of rural economy. Therefore, it is necessary to study the risk assessment system of farmers' small loans. This paper first introduces the concept of peasant household credit evaluation and the concept of farmers small credit loan, and then analyzes the domestic and foreign research on the risk assessment of farmers small credit loan in detail. Through analysis, it is found that the domestic research in this area is not as good as that in foreign countries. Artificial intelligence (AI) has been widely used in foreign countries because of its great advantage in the application of credit evaluation. BP neural network technology is an important technology in artificial intelligence technology, it has strong learning and adaptive ability, better internal parallel computing and storage, it is a stable nonlinear method and so on. Therefore, BP neural network is also applied in the risk assessment of farmers' small credit loan. However, the traditional BP neural network has some shortcomings, such as slow convergence rate and easy to fall into local minima. To solve these problems, many improved strategies have been put forward, such as adding momentum term, adaptive learning rate, LM algorithm, artificial immune algorithm, genetic algorithm, particle swarm optimization algorithm and so on. In this paper, a new swarm intelligence algorithm, quantum particle swarm optimization (QPSO), is used to improve the BP neural network model. Quantum Particle Swarm Optimization (QPSO) has the advantages of less adjusting parameters, simple and easy to realize, and has better convergence performance and global search ability. It can overcome the shortcoming of BP neural network algorithm in convergence performance to a certain extent. The quantum particle swarm optimization (QPSO) algorithm (QPSO) is improved by adding adaptive mutation, and good results are obtained. Because the quantum particle swarm optimization (QPSO) converges fast in the early stage, it has gathered to a certain point and formed a local minima when the global optimization is not reached in the later stage. An adaptive mutation is added to improve the quantum particle swarm optimization (QPSO) algorithm (QPSO). The improved Quantum Particle Swarm Optimization (QPSO) algorithm (QPSO) obtains the improved BP neural network model by optimizing the weights and thresholds of the BP neural network. Compared with the improved BP neural network model such as genetic algorithm, it can improve the convergence speed of BP neural network more effectively and prevent it from falling into local minima. Then, the improved BP neural network model is applied to the risk assessment system of farmer's small credit loan. In the simulation experiment, five groups of data sets are randomly selected from the data to carry out the experiment, and then the average values of the five groups of experimental results are compared with the traditional BP neural network. The improved BP neural network can improve the accuracy of credit evaluation and reduce the possibility of error. In the end, this paper looks forward to the next step from the aspects of the risk assessment of farmer's small credit loan and the improvement of BP neural network.
【学位授予单位】：安徽大学
【学位级别】：硕士
【学位授予年份】：2014
【分类号】：F832.4;TP183
本文编号：2304629
[Abstract]:With the rapid development of rural economy, rural credit loan demand also increases. Rural credit loans to farmers mainly small credit loans. The credit risk assessment of farmers is very important in farmers' small loans, which restricts the development of rural economy. Therefore, it is necessary to study the risk assessment system of farmers' small loans. This paper first introduces the concept of peasant household credit evaluation and the concept of farmers small credit loan, and then analyzes the domestic and foreign research on the risk assessment of farmers small credit loan in detail. Through analysis, it is found that the domestic research in this area is not as good as that in foreign countries. Artificial intelligence (AI) has been widely used in foreign countries because of its great advantage in the application of credit evaluation. BP neural network technology is an important technology in artificial intelligence technology, it has strong learning and adaptive ability, better internal parallel computing and storage, it is a stable nonlinear method and so on. Therefore, BP neural network is also applied in the risk assessment of farmers' small credit loan. However, the traditional BP neural network has some shortcomings, such as slow convergence rate and easy to fall into local minima. To solve these problems, many improved strategies have been put forward, such as adding momentum term, adaptive learning rate, LM algorithm, artificial immune algorithm, genetic algorithm, particle swarm optimization algorithm and so on. In this paper, a new swarm intelligence algorithm, quantum particle swarm optimization (QPSO), is used to improve the BP neural network model. Quantum Particle Swarm Optimization (QPSO) has the advantages of less adjusting parameters, simple and easy to realize, and has better convergence performance and global search ability. It can overcome the shortcoming of BP neural network algorithm in convergence performance to a certain extent. The quantum particle swarm optimization (QPSO) algorithm (QPSO) is improved by adding adaptive mutation, and good results are obtained. Because the quantum particle swarm optimization (QPSO) converges fast in the early stage, it has gathered to a certain point and formed a local minima when the global optimization is not reached in the later stage. An adaptive mutation is added to improve the quantum particle swarm optimization (QPSO) algorithm (QPSO). The improved Quantum Particle Swarm Optimization (QPSO) algorithm (QPSO) obtains the improved BP neural network model by optimizing the weights and thresholds of the BP neural network. Compared with the improved BP neural network model such as genetic algorithm, it can improve the convergence speed of BP neural network more effectively and prevent it from falling into local minima. Then, the improved BP neural network model is applied to the risk assessment system of farmer's small credit loan. In the simulation experiment, five groups of data sets are randomly selected from the data to carry out the experiment, and then the average values of the five groups of experimental results are compared with the traditional BP neural network. The improved BP neural network can improve the accuracy of credit evaluation and reduce the possibility of error. In the end, this paper looks forward to the next step from the aspects of the risk assessment of farmer's small credit loan and the improvement of BP neural network.
【学位授予单位】：安徽大学
【学位级别】：硕士
【学位授予年份】：2014
【分类号】：F832.4;TP183
【参考文献】
相关期刊论文 前8条
1 胡愈;许红莲;王雄;;农户小额信用贷款信用评级探究[J];财经理论与实践;2007年01期
2 杜晓山,孙若梅;中国小额信贷的实践和政策思考[J];财贸经济;2000年07期
3 谭民俊;王雄;岳意定;;FPR-UTAHP评价方法在农户小额信贷信用评级中的应用[J];系统工程;2007年05期
4 李文政;唐羽;;国内外小额信贷理论与实践研究综述[J];金融经济;2008年16期
5 曹道胜;何明升;;商业银行信用风险模型的比较及其借鉴[J];金融研究;2006年10期
6 熊铭奇;毛雅娟;;中国农村信用的特殊性及信用体系的构建[J];农村经济;2009年10期
7 孙清;汪祖杰;;LOGIT模型在小额农贷信用风险识别中的应用[J];南京审计学院学报;2006年03期
8 温涛,冉光和,王煜宇,熊德平;农户信用评估系统的设计与运用研究[J];运筹与管理;2004年04期
，本文编号：2304629
本文链接：https://www.wllwen.com/jingjilunwen/guojijinrong/2304629.html
