Integrating Nonlinear Dimensionality Reduction with Random Forests for Financial Distress Prediction

Abstract

Abstract With the recent financial crisis, developing accurate financial distress prediction models has become more important. Due to the high-dimensionality of the input data, this study proposes to integrate nonlinear dimensionality reduction (NLDR) techniques, such as isometric feature mapping (ISOMAP) and locally linear embedding (LLE) with random forests (RF) to develop a novel prediction for financial distress. These techniques help to reduce the dimensionality of input data and enhance the performance of RF classifiers. The effectiveness of this methodology has been verified by experiments that compare it to classical linear dimensionality reduction techniques. Empirical results indicated that our hybrid approach outperforms classical linear dimensionality reduction techniques with RF. Moreover, the ISOMAP has better performance than other dimensionality reduction techniques.