The financial distress causes serious damage to employees, creditors, shareholders and other stakeholders, as well as social and economic turmoil. Therefore, it is very important to establish an effective financial distress prediction model. The sample of this research is taken from the database of Taiwan Economic Journal (TEJ), using Taiwan listed OTC companies with financial distress from 2010 to 2020 as the sample object and matching 1:2, using 14 financial variables and 4 non-financial variables. First, the CHAID and the support vector machine are used to screen important variables, and then the variable data is input into the convolutional neural network for training, respectively, to establish a financial distress prediction model. The empirical results show that the model (SVM-CNN) established by using the support vector machine with the convolutional neural network is compared with the unfiltered convolutional neural network model and the CHAID with the convolutional neural network model , the convolutional neural network model screened by support vector machine is the best financial crisis prediction model in this study, and its average accuracy rate is 89.74%.
