Abstract: Illegal parade incidents seriously harm the public order. The accurate prediction of illegal parade paths will be guidable for the efficient deployment of police force. Existing related studies mostly do not consider the comprehensive influence of the attraction factors and the repulsion factors on path selection of parade groups, resulting in the low precision of the prediction of the parade movement path. Therefore, this paper proposes a multi-factor constrained dynamic prediction method of illegal parade paths. Specifically, a dynamic weighted undirected graph is firstly constructed by combining the real-time locations of the parade group with the spatial distributions of multiple factors based on the distance decay theory. Then, an optimization model is built with the constraints of multiple factors, and the Floyd algorithm is employed to predict the parade paths from the parade group’s current position to the destination. Comparative experiments on three historical illegal parades show that the matching degrees between the obtained results by the proposed method and the real parade paths are higher than 80% in terms of path length, road segments and road network nodes, demonstrating the efficiency and advan- 
tages of the proposed method.

Key words: illegal parade, path prediction, multi-factor constraints, distance decay, Floyd algorithm