Research On Early-warning Of Bank Risk Based On Support Vector Machine And Mobile Agent Technology

With the rapid development of banking industry, Banking Regulatory Commission pays more and more attention on the early warning of bank risk. Early warning system for banks is studied in this paper. Support vector machine technology and mobile agent technology are applied in the system. An early warning method for bank risk based on support vector machine classification (Support Vector Classification, SVC) is proposed in this paper. And an early warning method based on experts'experience and uncertain support vector machine classification is presented. The experiments prove that the above models and methods are superior to the traditional ones.Secondly, to the large amount of data in banks, centralized computing model consumes much time, and requires high level of hardware. Moreover, the network structure of the early warning system is very complicated, so the parallel computing of mobile agent technology is introduced in this paper. In the mobile agent technology framework, the mechanism of agent service registration and location, the establishment of migration route, the selection and update of route are studied in this paper. The contents are as follows;1 This paper analyzes the early warning of bank, pattern classification and support vector machines. Then look on banking early warning as a pattern classification process. Support vector machine is an effective way to solve classification problem. This paper proposes a model of bank warning based on support vector machine, realizes the automatic selection of relevant model parameters, and gives out the algorithm steps and the experiment analysis.2 To the uncertainty of bank warning system, this paper proposes an early warning method based on uncertainty support vector classification. This method takes the experts'experience and uncertainty information into account. It achieves a combination of early warning method and intelligence experts, and gives out the bank warning instance. It provides new ideas for the early warning system of bank risk.3 Because of a great amount of data in banks, centralized computing model is time-consuming, and demands high level hardware. Moreover, the network of banking system is wide distributed and the structure is complicated. So this paper introduces the mobile agent technology into early warning system. To the shortcoming of traditional routing algorithm for mobile agent, a new routing protocol based on ant colony is proposed. And it realizes real-time routing updating using the idea of feedback in automatic control theory. This paper proposes that the mobile agent takes the task of routing updating in the implementation of calculatiing. So it realizes the real-time routing information updating with a minimum consumption of bandwidth and delay, in order to meet the banking requirement of early warning systems.4 Traditional agent migration theory often leads to overcrowding on the network nodes of optimal path in the case of heavy network load. It can result in longer migration delay or even result in the loss of agent. Therefore, this paper proposes a migration algorithm based on probability, considering the bandwidth, time lapse and other factors. It calculates which route to selecte by computing the weight of each routes, which changes the feature that agents only migrate in accordance with the optimal path in the past. So every possible route can be reasonable utilized and it can alleviate the pressure on the best route effectively. Then it achieves good network load balancing of the bank early warning system.5 Given the current structure of the network system, traditional agent services for searching need large bandwidth and long time lapse. Such structure is not suitable for strong mobility and effectiveness, so this paper presents a new mobile agent service registration algorithm based on regional center. It can overcome the shortcomings of long search time and slow updating.