In recent years, newly-developed data mining and machine learning techniques have been applied to various fields to build intelligent information systems. However, few of these approaches offer online support or are able to flexibly adapt to large and complex financial datasets. Therefore, the present research adopts particle swarm optimization (PSO) techniques to obtain appropriate parameter settings for subtractive clustering (SC) and integrates the adaptive-network-based fuzzy inference system (ANFIS) model to construct a model for predicting business failures. Experiments were conducted based on an initial sample of 160 electronics companies listed on the Taiwan Stock Exchange Corporation (TSEC). Experimental results show that the proposed model is superior to other models, providing a lower mean absolute percentage error (MAPE) and root mean squared error (RMSE). The proposed one-order momentum method is able to learn quickly through one-pass training and provides high-accuracy short-term predictions, while the proposed two-order momentum provides high-accuracy long-term predictions from large financial datasets. Therefore, the proposed approach fulfills some important characteristics of the proposed model: the one-order momentum method is suitable for online learning and the two-order momentum method is suitable for incremental learning. Thus, the PS-ANFIS approach could provide better results in predicting potential financial distress.
All Science Journal Classification (ASJC) codes