Financial Big Data Analysis by Sparse Representation Classifiers

Financial big data analysis has recently become a popular research field. Kernel machines (such as support vector machines, SVM) have demonstrated good performance in many areas of pattern recognition. However, the representation of traditional kernel machines is not sparse. A sparse model representation in machine learning is expected to improve the generalization performance and computational efficiency. Moreover, in big data analysis, high-dimensional and nonlinear distributed data generally degrade the performance of a classifier due to the curse of dimensionality, especially in financial distress predictions. To address these problems, this study proposes a novel system using kernel sparse representation classifiers (KSRC) to discriminate financial statement data. The statement data are first projected to a low-dimensional subspace and is then classified by the KSRC. Compared with other data mining systems, the proposed system performs best.