オンライン人流予測

抄録

In this paper, we consider online people-flow prediction problem. People flow is influenced by various environmental factors such as behavior characteristics and seasonal changes that make it difficult to solve the problem by statistical analysis. The approach presented in the paper is based on online prediction algorithms, in which the predictor assumes an adversarial environment. To consider adversarial settings, we can guarantee the performance in any environment. We show the experimental results of congestion prediction at some locations in our university.