Abstract: In probabilistic prediction PRoPHET routing the effectiveness of the forwarding node was not considered. There are disadvantages that in the case of the same predictability of the multiple relay nodes, mobility will cause the connection time between nodes to be different, resulting in the low message delivery rate and excessive network load. Therefore, the influence of the node encounter duration on the complete transmission of message data should be considered when calculating the prediction probabilities. This paper proposes a probabilistic routing algorithm based on Connection Time (PRoPHET-CT), which incorporates the factors that influence the duty cycle of the connection time between nodes. The comparison experimental results show that the improved algorithm has a higher probability of delivery in the network environment with high node density. And also, this algorithm can reduce the network load and the average hops of the nodes.