In opportunistic networks, data communication is achieved through opportunistic contact between nodes. Due to the limited availability of resources, nodes may exhibit selfish behavior during data transmission and be unwilling to collaborate on forwarding tasks. Therefore, it is necessary to adopt certain incentive mechanisms to encourage user participation. However, most existing incentive mechanisms fail to account for games under conditions of incomplete information, nor do they address the risks associated with potential breaches of node privacy, which could be exploited by malicious actors, thereby posing significant security risks. Therefore, in this paper, a Deep Reinforcement Learning-based Dynamic Incentive Mechanism, named DRL-DIM, is proposed that constructs node interactions as a Stackelberg game model. In the context of unknown priors, game incentives are transformed into a learning decision problem, the deep reinforcement learning algorithm PPO is introduced for learning and training, which facilitates the obtain of optimal node strategies through dynamic learning, thereby promoting the achievement of game theory balance and motivating nodes to actively participate in data transmission. Simulation experiments have been conducted on a large number of real datasets, the results show that, our proposed mechanism demonstrates better performance in terms of delivery ratio and average end-to-end delay compared to the contrasting incentive mechanisms.