Credit scoring model based on PCA and improved tree augmented Bayesian Classification

According to the features of high dimensional, nonlinear and redundant of the Credit Scoring data, the establishment of a model for credit scoring has a direct bearing on the complexity of personal credit scoring process and the collection of characteristic parameters reflecting the credit scoring status constitutes an important link for setting up a efficient model,to resolve the problem that it is difficult to reduce the dimension and the classification accuracy rate is low in traditional methods, a novel Credit Scoring model is proposed based on Principal Component Analysis and improved tree augmented Bayesian Classification. It first uses principal component analysis to eliminate redundant information and simplify the Bayesian network´s inputs. Then establishes an improved tree augmented Bayesian Classification models for personal credit scoring. The algorithms have been validated experimentally by using real data. Theoretical and experimental results show a performance competitive with the state-of-the-art and a higher classification accuracy.