The aim of change mining is to discover changes in process models based on execution data recorded in event logs. There may be hidden transitions in the process models related to, for example, business integration and user requirements that do not exist in event logs. Behavioral change mining in the case of hidden transitions is a fundamental problem in the field of change mining. Existing research on change mining has not considered the effects of hidden transitions. This paper proposes a novel method based on complete logs with hidden transitions for mining behavioral changes. We analyze the behavioral relations of activities based on changed logs under the condition that the original model is unknown. Log-driven change mining is realized by calculating the log behavioral profile, minimum successor relation, and log-weighted coefficient, which allows the mining of hidden transitions, as well as changed behavioral relations. Finally, this method is applied to disaster chain risk analysis, and the evolution of disaster chains in different scenarios is mined from disaster logs to determine the type of disaster chain. The results of this paper provide a scientific basis for the strategy of chain-cutting disaster mitigation in the emergency management of disaster chains.