Flooding is the third biggest disaster in the world according to the World Meteorological Organization. Several methods like numerical models, physical models, and Machine Learning (ML) models have been engaged in flood prediction to minimize the impact of flooding. Despite the improvements experienced in the use of some ML methods, there are still drawbacks due to accuracy. Hence, this study evaluated Categorical Boosting Algorithm (CatBoost) for flood prediction based on some evaluation metrics. Relevant flood-predictive factors were identified from the Osun River basin. The data was split into 70% for the training and 30% for the testing of the algorithm. The flood dataset was imported into the CatBoost Algorithm using Python programming language with the default parameters of the algorithm. The algorithm was evaluated using accuracy, precision, sensitivity, and multiclass loss function. The results showed that the accuracy, precision, and sensitivity of the CatBoost Algorithm were 92.48%, 63.82%, and 85.86% respectively. The result of the multiclass loss function during validation was 0.165874, which was significantly lower than the result during training, which was 0.925104. This indicates that the algorithm is overfitting the training data and is not generalizing well to new data. This can be a prospect for further study.