This paper applies three machine learning algorithms, namely decision tree, random forest, and AdaBoost, and two hybrid algorithms, particle swarm optimization and genetic algorithm, to monthly water prediction data. Experiments were carried out on the train and test set according to the parameters affecting the performance of the relevant algorithms. Further, the implementations of the performed algorithms are experimentally compared with each other in the training and testing stage by providing graphical illustrations of the İstanbul water consumption dataset. The numerical results indicate that the random forest algorithm has shown very decent results in the training and testing phase by providing the 0.92 R2 and 0.0238 mean absolute percentage error (MAPE) and 0.1493 MAPE and 0.83251 R2 respectively.