Accurately predicting urban runoff plays a crucial role in flood risk management; however, this remains a challenging task due to its unpredictable and uncertain nature. To ensure precise predictions, it is essential to identify the most significant influencing variables. Therefore, the primary objective of this research is to employ an artificial neural network (ANN) to pinpoint the most relevant parameters for monthly runoff prediction. For this study, four input variables are considered: maximum and minimum temperatures, rainfall, and soil moisture. The city of Mogadishu, Somalia, serves as the case study location. Global meteorological data from the period between 1985 and 2022 are collected for analysis. The findings of this research indicate that rainfall and soil moisture are the most critical input parameters, enabling greater accuracy and reducing complexity in predicting runoff. These two factors directly and significantly impact the quantity and behavior of water flow over the land surface, making them pivotal in determining runoff patterns.