Structuring the relationships through which events occur is crucial for gaining a deeper understanding of the past and for applying that knowledge to the present. In this study, we propose a method for constructing event networks that represent dependencies between events using large language models (LLMs). Our method performs both network and node selection to identify event pairs prior to applying the LLM. To evaluate the method, we collected past event data from Wikipedia and created a network dataset. An evaluation of our method showed an F1 score of 45.1%, demonstrating its effectiveness.