An intelligent credit forecasting system using supervised nonlinear dimensionality reductions

Kernel classifiers (such as support vector machines) have been successfully applied in numerous areas, and have demonstrated excellent performance. However, due to the high dimensionality and nonlinear distribution of financial input data in credit rating forecasting, finding a suitable low dimensional subspace by nonlinear dimensionality reductions is a key step to improve classifier performance. By integrating supervised kernel locality preserving projections (SKLPP) with kernel classifiers, this study develops a novel forecasting system for credit ratings. SKLPP is employed to gain a perfect approximation of data manifold and simultaneously preserve local within-class geometric structures according to prior class-label information. Empirical results indicate that, compared with other dimensionality reduction methods, the performance improvement owing to SKLPP is significant. Moreover, the proposed hybrid classifier outperforms other conventional classifiers.