This research was conducted with the aim of predicting the risk of bankruptcy of banks listed on the Iranian and Iraqi stock exchanges during the Covid-19 era, using neural network and machine learning approaches. The study covers two time periods: the pre-Covid-19 era (2003 2019) and the post-Covid-19 financial crisis era (2023 2020). The main objective of the research is to specify a model to predict the bankruptcy of Iranian and Iraqi banks. Different machine learning models including decision tree, XGBoost and random forest were used. The results showed that the most important variables in predicting bank failure risk include LTICA, ILQ, LRG, CLRE, NIRBT, NIRAT, NPS, ELTL, GP, ETA, NIRA, POAI, IR, NLTA, OPR, ROA, GDP and CYCLE. Therefore, a comparison of the periods before and after the Covid-19 crisis using neural networks and support vector machines (SVMs) showed that the accuracy of the models increased during the crisis period. The predictive accuracy of the SVM model during the Covid-19 crisis reached 0.9321, which increased compared to the pre-crisis period (0.9125). There was also an improvement in the sensitivity and specificity indices for all three classes during the crisis period .This study shows the importance of using multiple approaches to predict bank failure risk. The combination of different machine learning models and neural networks allows a more accurate identification of the factors that influence the risk of failure. The important role of macroeconomic factors such as GDP and interest rates in predicting bank failure risk was also identified. The results of this research highlight the importance of government support measures during the crisis and show that banks' bankruptcy risk management performance has improved during the Covid 19 era. The study provides practical suggestions, including the development of an early warning system, improvement of risk management systems, and regular training programmes for bank employees and supervisors.