This research has investigated the comparative role of bank credits and macroeconomic variables in predicting financial crises in Iran and Iraq by using machine learning approaches, neural networks and economic data of the two countries during the period 2000 to 2023. Various models for predicting financial crises (FC) using yield curve slope (YCS), debt service ratio (DSR), consumer price index (CPI), investment (INV), current account (CA), public debt (PD) variables. and bank credits (BC) were developed .The results indicate that in both countries, YCS and CA have the most influence in predicting financial crises. In Iran, the fuzzy system model showed the best performance with 79.18% accuracy and in Iraq, the cerebellar neural network with 67.89% accuracy. The optimal expectation algorithm showed that YCS is the most important predictive variable with a significance of 49.26% in Iran and 36.53% in Iraq. In contrast, BC was relatively less important in both countries (2.63% in Iran and 2.07% in Iraq). Therefore; The analysis showed that the average probability of financial crisis for Iran (0.9679) is slightly higher than Iraq (0.925), which can indicate a higher risk of financial crisis in Iran. This study reveals that despite the similarities in the factors affecting financial crises in the two countries, the relative importance of these factors is different. These findings emphasize the necessity of adopting distinct approaches and appropriate to the specific conditions of each country in economic policy-making and highlight the importance of paying attention to the unique economic characteristics of each country in the design of financial crisis prevention strategies.