Floods devastate communities, inflicting heavy economic and human losses. Traditional forecasting methods often struggle to account for the complex interplay of factors contributing to floods. This research introduces a hybrid machine learning algorithm designed to improve performance in flood forecasting for urban areas. This hybrid CSO-ELM model employs a swarm intelligence algorithm known as cuckoo search optimization (CSO) to optimize the number of hidden layer nodes in extreme learning machine (ELM). This approach keeps the core attribute of ELM whilst improving the performance of the model for flood forecasting. Application of the model on a case study data set of Kuala Lumpur, Malaysia shows that it consistently performs better in evaluation metrics used. Its architecture which combines the strengths of CSO and ELM algorithms overcomes the limitations of traditional approaches and reduces the need for manual parameter tuning.