Development and application of a hybrid forecasting framework based on improved extreme learning machine for enterprise financing risk

A scientific framework that can effectively forecast enterprise financing risks can both promote enterprise management and reduce the cost of risk for financial institutions. This study constructs a novel hybrid forecasting framework for enterprise financing risk incorporating modules for data preprocessing, feature selection, fore-casting, and evaluation. Specifically, the data preprocessing module mainly realizes the prescreen financing risk indicators and solves the forecasting challenge created by imbalanced data; The feature selection module based on binary grey wolf optimization is designed to intelligently identify optimal financing risk indicators; The forecasting module based on the improved extreme learning machine model established in this paper achieves higher forecasting accuracy; and the evaluation module provides reasonable and scientific evaluations of the proposed hybrid forecasting framework by using the data from small and medium-sized enterprises (SMEs) in China and all listed enterprises with Shanghai and Shenzhen A-shares. Using the SMEs dataset as an example, the Type-2 error value of the developed hybrid forecasting framework is 0.1765, which is 70.24% lower than the average result of the other models; the G-mean value of the framework is 0.8566, which is 40.56% higher than the average result of the other models. Based on the results, the proposed hybrid forecasting framework out-performs other comparative models and is a reliable tool for forecasting enterprise financing risk.