Heatstroke is becoming an increasingly serious threat to outdoor activities, especially, at the time of large events organized during summer, including Olympic Games or various types of happenings in amusement parks like Disneyland or other popular venues. The risk of heatstroke is naturally affected by a high temperature, but it is also dependent on various other contextual factors such as the presence of shaded areas along traveling routes or the distribution of relief stations. This study proposes a framework to evaluate the heatstroke risk in geographical areas by utilizing context-aware indicators which are determined by large and heterogeneous data including facilities, road networks and street view images. Based on the evaluation metric of the total heatstroke risk in the target area, we propose a Mixed Integer Nonlinear Programming model for optimizing routes of pedestrians, determining the location of relief stations and the supply volume in each relief station. Our experiments conducted on the planned site of the Tokyo Olympics and simulated during the two weeks of the Olympics schedule indicate that planning routes and setting relief stations with our proposed optimization model could effectively reduce heatstroke risk, thus letting organizers better prepare for the event.