Linear versus nonlinear dimensionality reduction for banks’ credit rating prediction

作者：

Highlights：

• Linear and nonlinear dimensionality reduction are applied for banks’ rating prediction.

• PCA and a variant of Isomap are compared by means of extensive computational tests.

• Dimensionality reduction induced an improvement in the classification accuracy.

• Isomap-based method outperformed PCA by providing more accurate predictions.

摘要

•Linear and nonlinear dimensionality reduction are applied for banks’ rating prediction.•PCA and a variant of Isomap are compared by means of extensive computational tests.•Dimensionality reduction induced an improvement in the classification accuracy.•Isomap-based method outperformed PCA by providing more accurate predictions.

论文关键词：Dimensionality reduction,Manifold learning,Isometric feature mapping,Banks’ credit rating prediction,Multi-category classification

论文评审过程：Received 4 September 2012, Revised 23 January 2013, Accepted 1 March 2013, Available online 13 March 2013.

论文官网地址：https://doi.org/10.1016/j.knosys.2013.03.001