Due to the explosive growth of mobile data traffic, it has become a common practice for Mobile Network Operators (MNOs, also known as operators or carriers) to utilize cellular and WiFi resources simultaneously through mobile data offloading. However, existing offloading technologies are mainly established between operators and third-party WiFi resources, which cannot reflect users dynamic traffic demands. Therefore, MNOs have to design an effective incentive framework, encouraging users to reveal their valuations on resources. In this paper, we propose a novel bid-based Heterogeneous Resources Allocation (HRA) framework. It can enable operators to efficiently utilize both cellular and operator-own WiFi resources simultaneously, where the decision cost of user is strictly controlled. Through auction-based mechanisms, it can achieve dynamic offloading with awareness of users valuations. And the operator-domain offloading effectively avoids anarchy brought by users selfishness and lack of information. More specifically, HRA-Profit and HRA-Utility, are proposed to achieve the maximal profit and social utility, respectively. addition, based on Stochastic Multi-Armed Bandit model, the newly proposed HRA-UCB-Profit and HRA-UCB-Utility are able to gain near-optimal profit and social utility under incomplete user context information. All mechanisms have been proven to be truthful and satisfy individual rationality, while the achieved profit of our mechanism is within a bounded difference from the optimal profit. In addition, the trace-based simulations and evaluations have demonstrated that HRA-Profit and HRA-Utility increase the profit and social utility by up to 40% and 47%, respectively, compared with benchmarks. And the cellular utilization rate is kept at a favorable level under the proposed mechanisms. HRA-UCB-Profit and HRA-UCB-Utility restrict pseudo-regret ratios under 20%.