Search

Credit scoring is explored to assess default risk of consumer behaviors for financial institutions, banks in particular. The advanced Bayesian algorithm is proposed for credit assessment. The new trial ensembles logistic regression analysis (LRA), cluster and MLP-NN in Bayesian approach as an advanced classifier. The investigation contain evidence that Bayesian ensemble technique optimizes LRA, cluster and MLP-NN consequence in credit assessment. Simultaneously, Bayesian approach creates positive advantage on algorithm accuracy and fitness for credit evaluation, nevertheless, balances and decreases classified error ratio (ER).

With the rapid development of China's consumer credit market, personal credit scoring has attracted more and more attention. And a variety of statistical and artificial intelligence methods have been used in personal credit scoring. This paper presents a multiple classifier fusion method, namely using relevant theory and methods to combine single classifiers. This is for making full use of the complementary information between classifiers, and ultimately achieves the purpose of better classification results. We present an investigation into the fusion of two different classification methods for personal credit scoring, using Dempster-Shafer's rule of fusion. These methods include traditional linear Logistic Regression and newly developing nonlinear BP neural network. Our experiment's results show that the performance of the fusion of these two different classifiers on real consumer credit customer data is good. Its classification accuracy is better than that of the individual method and its type II error rate is effectively reduced, taking the advantage of the fusion of two classifiers. And it exerts a very important significance on the control of the risk of commercial banks' business.

With the escalation of the industrial structure, what the future of the Yangtze River Delta manufacturing will be, how the Yangtze River Delta to adjust the industrial structure to maintain good economic growth has become a community of common concern. This paper first uses principal component analysis (PCA) to identify the leading industry of the Yangtze River delta, and chooses three representative industries, namely the black metal smelting and rolling processing industry, general equipment manufacturing and computer and communication equipment manufacturing industry, to analyze the dynamic variation characteristic of the gravity centre of the leading industries from variation direction, distance by building industrial gravity center model. At last, we contrastively summarize the dynamic variation characteristic of the gravity centre of the leading industry of the Yangtze River Delta.

In this paper, we study the synchronization problem for singular complex networks with time-varying delays by Takagi-Sugeno(T-S) fuzzy theory. A modified Lyapunov-Krasovskii function is used to derive a sufficient condition for synchronization in terms of LMIs. A example is given to support the main result of this paper.

In this paper, the exponential topology identification for the uncertain complex dynamical network with time-varying coupling delay is studied via periodically intermittent control for the first time. By designing an periodically intermittent feedback controller, the unknown network topological structure is identified in the process of synchronization. Finally, an illustrative example is given to show the effectiveness of the theoretical results.

Although industrial cluster attracts great attention, it still holds some research deficiencies. Rare exploration about cluster's external linkage is the most significant one. To make up such deficiency, this paper takes hub-spoke industrial cluster as example to analyze the influence and effect path of external linkage on cluster. Finally for more information about external linkage's effects, this paper identifies the order of different external linkage forms with ZhongGuanCun(ZGC) as research object and further analyzes the reason of the final quantitive result.

This paper proposes an approach of part synchronization to identify the unknown parameters of the different structures of uncertain chaotic system. Based on Asymptotic stability theory, we design an observer to identify the unknown parameters in the uncertain chaotic system. At the same time, in order to achieve the different structure of uncertain Lorenz chaotic system and Coullet synchronization, we construct the synchronous controller and determine its gain range by using the backstepping approach. The simulation results demonstrate the effectiveness of the observer and controller. In comparison with the existing literature, we optimize the observer and the controller which can achieve the part synchronization of the different structures of uncertain chaotic systems.

This paper investigates out synchronization with identical topologies connection and the adaptive synchronization with nonidentical topologies connection between two nonlinearly delay-coupled complex networks with the bidirectional actions. A new Lyapunov-Krasovskii functional is constructed for the two complex networks, and the S-procedure is employed to handle the nonlinearities and identical topologies networks. Based on LaSalle invariance principle, combining the nonlinear adaptive controllers and update laws, some criteria are derived to ensure the synchronization between two complex networks with nonidentical topologies. Finally, a numerical example and its simulations are given to illustrate the efficiency of the derived result.

in the area of personal credit risk management business, the domestic issues haven't developed a personal credit scoring method which matches with the development of personal credit. To some extent this problem has constrained and impeded the sound development of China's consumer credit business. Therefore, based on the principles of combination forecast, this paper proposes optimization in the weight of a single model in combining models using SA-GA and constructs an individual credit score combination forecast model based on SA-GA. The result shows that the combination model can be effectively integrated with the advantages of a single model, and it also has more advantages in prediction accuracy and stability. Moreover, this model can effectively explain the impact of variables upon default, and shows relatively strong practical value.

In this paper, we investigate a n-patches nonautonomous predator-prey system with nonlinear diffusion. By the comparison theorem, the sufficient conditions are derived for uniform persistence of the system. We also prove the existence of positive periodic solution of the system by using Schauder's fixed point theorem. Furthermore, the sufficient conditions for global asymptotical stability of positive periodic solution are obtained by constructing an appropriate Lyapunov function. The conditions we obtained are more general. We also give the ecology interpretation to our results.

we propose Fuzzy ART in personal credit area to the problem which can’t handle discrete variables and continuous variables together and compared the outputs with logistic regression, linear programming and the BP neural network model results. The empirical result indicates that the model of Fuzzy ART category has less error II and possesses better applicable for commercial banks.

In this paper, the global asymptotic stability for a class of uncertain delayed cellular neural networks with discrete and distributed time-varying delays (DCNNs) is considered. Based on the Lyapunov functional stability analysis for differential equations and the linear matrix inequality (LMI) optimization approach, a new criterion is derived to guarantee global asymptotic stability. A numerical example is illustrated to show the effectiveness of our results.

The paper is mainly concerned with the issues of synchronization dynamics of complex dynamical networks with coupling delays and multi-links. A general model of complex networks with multi-links is formulated, which can well describe practical architectures of more realistic complex networks related to multi-links. Based on Lyapunov stability theory and adaptive control techniques, some simple controllers with updated feedback strength are introduced to make the network locally and globally synchronized. Furthermore, the effectiveness of theoretical results are verified by numerical simulations.

In this paper, the problem of robust adaptive convergence for uncertain Cohen-Grossberg neural networks(CGNNs) with mixed delays is investigated. Using the Lyapunov method and employing a novel lemma, some delay-independent conditions are derived to ensure the state variables of the discussed robust system to converge, globally, uniformly, exponentially to a ball in the state space with a pre-specified convergence rate. The effectiveness and usefulness of the results has been verified by a numerical example with graphical illustrations.

Using Chebyshev inequality and nonnegative semi-martingale convergence theorem, the paper investigates asymptotic behavior of stochastic Cohen-Grossberg neural networks with delay by constructing suitable Lyapunov functional. Algebraic criteria are given for stochastic ultimate bounded and almost exponential stability. The result in the paper extend the main conclusion. In the end, examples are given to verify the effective of our results.

Aiming at the insufficiencies of BP neural network, this paper established a hybrid neural network based on the combination of GA and BP algorithms. The hybrid algorithm made fully use of GA's global searching to improve the learning ability of neural network with the combination of BP. The model was used in personal credit scoring in commercial banks. Compared with single BP neural network, the training results of hybrid neural network indicate that the hybrid algorithm can improve the learning ability of neural network to achieve the training goal. The classification accuracy of hybrid neural network on testing samples is higher than that of single BP neural network.

Personal credit scoring plays an important role for commercial banks to keep away from consumer credit risks. This paper used neural networks for personal credit scoring and used two evolutional algorithms of genetic algorithm (GA) and particle swarm optimization (PSO) to train the networks to construct a GA neural network and a PSO neural network respectively. The two neural networks were used to classify the consumer credit data of commercial banks. Compared with BP neural network, the results indicate that GA network and PSO network get lower accuracies on training samples, but on testing samples, the accuracies of GA network and PSO network are higher than that of BP network by 0.38% and 0.76% respectively. On modelpsilas robustness, the accuracy differences on the two groups of samples of GA network and PSO network are lower than that of BP network by 2.08% and 1.33% respectively, which indicate that GA neural network and PSO neural network get a better robustness.

Aiming at low predictive accuracies of single models, this paper presents a combining forecast for personal credit scoring. Based on two single statistical models of linear regression and logistic regression, this paper constructed a non-linear combining forecast by genetic programming (GP) and used the constructed model for personal credit scoring. The application results indicate that the predictive accuracy of the non-linear combining forecast based on GP is higher than linear regression, logistic regression and the linear combining forecast based on least square method by 3.40%, 2.83% and 2.64% respectively. The non-linear combining forecast also gets a much lower type II error rate which is more significant for commercial banks to keep away from consumer credit risks.