Knowledge graphs (KGs) are inherently incomplete and require continuous maintenance to keep up with the latest status of the target domain. Various studies have been conducted to automate this maintenance, but it still relies heavily on human labor. To alleviate such a burden, predicting updates in KGs is useful for maintaining KGs. In this study, we focus on predicting which entities to be updated in KGs containing many entities. Our basic idea is to exploit the KG’s structural information and the edit history, making it possible to predict entities likely to be updated in the next time interval. We have constructed a dataset including edit history based on Wikidata, a general-purpose knowledge graph, and developed a method to predict entities that will be updated based on link relations and edit history, and confirmed their usefulness through experiments.