Model Study On Credit Scoring Of Registered Constructor

In order to overcome the weakness of traditional methods on credit scoring of registered constructor, such as qualitative analysis only, the lack of scientific credit scoring index system and model, this dissertation attempts to establish the index system on credit scoring of registered constructor and the credit scoring model based on Support Vector Machine (SVM) to improve the method of registered constructor credit scoring.This dissertation puts forward that practice credit index and individual credit index constitute the credit scoring index system of registered constructor, so as to assure the index system is comprehensive and scientific. Thereamong, practice credit index includes good behavior record accreditation standard and bad behavior record accreditation standard. The latter is established by the coding regulation of Ministry of Housing and Urban-Rural Development of the People's Republic of China, and the former is the forceful supplement of the index system.SVM is firstly applied to build the model on credit scoring of registered constructor. This dissertation adopts LIBSVM2.86 as the major modeling tool and uses grid-search technique and 5-fold cross-validation to select the optimal parameter values of various kernel functions. By means of the comparison of four kernel functions about accuracy of 5-fold cross-validation, the RBF kernel function is found out to be the most rational function, which supports the excellence of it.