Recent years have seen a significant growth in Internet of Things (IoT) technology consisting of a large number of devices embedded with sensors and deployed to perform monitoring and actuation tasks. The IoT devices collect large volumes of data that is usually uploaded to cloud to perform analytics and predictions. One of the main challenges in IoT is the transportation of large-scale data collected over a period of time at a remote site. Cellular networks are already facing explosive growth of mobile data traffic due to the proliferation of smart devices and traffic-intensive applications. An alternate solution is to perform the data offloading, where a portion of data can be offloaded from primary links and transferred using opportunistic terminal-to-terminal (T2T) network that relies on direct communication between mobile users, without any need for an infrastructure backbone. However, such approach may lead to data loss and delay if dynamics of time-varying topology and mobility of nodes is not taken care of. To address this challenge, we propose three prediction-based offloading schemes that exploit the mobility patterns and temporal contacts of nodes to predict future data transfer opportunities. We have utilized the High-level Petri Nets to model and formally analyzed the communication processes of the proposed schemes. The new symbolic model verifier (NuSMV) has been employed for the verification of the three schemes against the identified constraints. The verification results affirm the correctness and scalability of the models. The protocols are evaluated with performance metrics, such as the delivery ratio, latency, and overhead. Our results indicate significant improvement in performance compared to existing approaches. © 2019, The Author(s).