The Research On Theory And Application Of Credit Scoring

The nature of the market is exchange.The most important character of market economy is trusting each other on the traders’promise.Because of the exchange cost and information asymmetry, it is easy to lead the party with information superiority to trigger moral hazard through default,simultaneously likely to cause the party with information inferior position to make adverse selection. In this instance, the market order will be seriously disrupted, the result is the termination of normal trade and economic development.Credit is the guarantee of the contract fulfillment under modern market economy conditions, the credit problems of the market economy is the main issue of contract enforcement.The credit relationship mainly maintained by the contract,but incomplete contracts and imperfect credit constraint mechanisms in the implementation process of the contract are sufficient conditions for leading to default risk.Through joining the third-party monitoring mechanism such as the penalty cost, laws and regulations, discovering skills it can be proved that they can encourage all parties of the game to keep the faithful strategy choice. Credit scoring is such a non-mandatory third-party "soft" constraint mechanism that can be used to the daily process of economic exchanges.Since the90s of the20th century, with the accelerated process of domestic market economy, personal credit market develop very quickly, credit consumption and investment continued to expand, but the concomitant credit risks become more and more obvious. As our personal credit system’s construction is in the initial stage, the collection and use of personal credit information, the establishment of monitoring mechanisms and the management skills of credit risks are still very poor, especially lacking a set of scientific and effective and accurate scoring method. This dissertation will do some work and attempt relying on the basis of current situation of our personal credit information database.The path of this research follows this way. Firstly, review evolution process of the credit mechanism and economic theory.The next step analyzing contract, credit game theory evolution route and organic connection, thorough analysis triple relation and effect mechanism, looking for inner and external factors that can influence the game strategies. And then expounds the basic principle of credit scoring,metric factors and application scope, and apply to the government supervision areas,then discuss how to use the data from "Credit Information System of Hunan Province"to install credit scoring models of self-employed people..Under actual database conditions, after further classification, accession, comparison, analysis, judgment and the process of preparation of such processing, collecting and screening a set of objective comprehensive and scientific index that can fully reflect credit factors such as the personal moral conduct, payment ability, performance status, rewards and punishment records.After obtaining the data of self-employed people, this article try to use different methods to establish a "customized" credit scoring model that have practical application value. Meanwhile, analyse the precision of different models, compare their advantages and disadvantages; Through empirical analysis we want to find credit factors and correlativity for our business administrative supervision department to forecast credit risk of self-employed people. They can take earlier regulatory measures to prevent risk and establish a trustworthy mechanism of incentives and penalizing bad decision-making reference.Finally, summarize the contents of this paper, indicate the direction of future research and development trend.This paper attempts to use registration information data of self-employed people, after using statistics and non-statistical methods to establish credit scoring model, the results showes the Logistic regression has better stability and accuracy, among artificial intelligence methods SVM has advantage for wide application. Ultimately we determine the model that has highest precise degree and design the credit scoring card. Through the analysis of test samples, the distinction degree achieve a satisfactory level.It shows that the model has very strong application value.This research aims at the largest number of economic activity principals, the results have great significance for supervised classification their credit activities and conditions.