This study proposes a hot water demand prediction method for operational planning of polymer electrolyte fuel cell cogeneration systems (PEFC-CGSs). PEFC-CGSs provide hot water by utilizing waste heat produced in the electricity generation process. An optimal operational plan according to household demand leads to further energy saving. Therefore, operational planning methods based on household demand prediction have received intense focus. In particular, the prediction of the amount of hot water demand is important for efficient operation. The authors have attempted to improve the hot water prediction method based on multivariate random forest (MRF), which uses the average of many decision trees' outputs as the prediction result. However, some experimental results show that a prediction strategy based on averaging the outputs of decision trees does not always lead to the best solution. In this study, the authors propose a novel prediction method utilizing the quantile of the estimation results derived in MRF. By setting the appropriate quantile, we can evade the demand underestimation, which has a higher negative impact on operational efficiency than overestimation. The usefulness of the proposed approach is evaluated via numerical simulations using real-world demand data.