<p>To address the issues of low accuracy and high misclassification rates in financial distress prediction (FDP) models, this study proposes a classifier model based on a backpropagation algorithm and a discriminant-restricted Boltzmann machine (BP-DRBM). Furthermore, a two-stage selective ensemble approach, employing stepwise selection and majority voting methods, is implemented to enhance model performance. An empirical analysis is conducted using data from 1,440 Chinese listed companies from 2018 to 2023. The results indicate that: (1) the BP-DRBM model achieves an accuracy of 85.7%, an improvement of 8.4% over the DRBM model; (2) In the stage I of selective ensemble, the total accuracy of the three ensemble models are 93.3%, 91.7%, and 91.7%, all significantly outperforming the individual base classifiers; (3) After two-stage selective ensemble, the overall accuracy of the model increases to 95.0%, representing an improvement of 11.2% over the baseline model; (4) Regarding misclassification rates, the average misclassification rate for positive samples is 8.6% after the stage I of ensemble, and it decreases to 3.7% after the stage II. These results demonstrate that the two-stage selective ensemble method based on BP-DRBM effectively improves the accuracy of FDP model while significantly reducing the misclassification rate. This study introduces a novel approach for FDP research.</p>