The artificial bee colony algorithm has the advantages of a simple structure and easy implementation. However, due to its strong exploration ability and weak exploitation capability, it also suffers from drawbacks such as low solution accuracy and slow convergence speed. To solve these problems, this paper proposes a dual search mode artificial bee colony algorithm based on exploration-exploitation tradeoff, which includes three core modules: exploration-exploitation control, exploration-exploitation execution, and exploration-exploitation reinforcement. In the control stage, we design an exploration-exploitation indicator based on the population's evolutionary dynamics, which initially emphasizes exploration and gradually shifts towards exploitation. In the execution stage, we design search equations guided by differential solution information, with a focus on the implementation points of exploration and exploitation. In the enhancement stage, the onlooker bees are instructed to select high-quality food sources for further search, using criteria such as diversity ranking and fitness ranking. Numerical experiments are conducted on traditional test functions and CEC2013 test functions. The results demonstrate that compared with eight high-level artificial bee colony algorithms proposed in recent years, the proposed algorithm exhibits strong competitiveness in terms of solution quality and convergence speed. Finally, the proposed algorithm is applied to optimize the risk prediction model for esophageal cancer, yielding satisfactory results.