The present study was conducted with the aim of flood risk zoning in Nekarud. For this purpose, some important variables were prepared using multi-source data and standardized in a GIS environment. Random forest, support vector machine and multilayer perceptron machine learning algorithms were used for modeling and their performance was evaluated with indicators such as overall accuracy, F1 score and area under the curve. The results showed that the multilayer perceptron model had the best performance with an accuracy of 94.5% and an area under the curve of 91.11%, while the random forest model with an accuracy of 88.64% and an area under the curve of 88.64% and the support vector machine model with an accuracy of 87.2% and an area under the curve of 82.92% were ranked next. In terms of spatial distribution, the random forest model assigned the highest contribution to the medium and low flood risk classes, the support vector machine model predicted a higher contribution to high-risk areas, and the multilayer perceptron model provided a balanced distribution. Analysis of the significance of variables in the random forest model showed that distance from the river, slope, and topographic moisture index played the greatest role in flood occurrence.