Although many patent recommendation methods have been proposed to suggest suitable patents, they aim to meet the technological needs of individual companies. Identifying the common technological needs of companies in an industrial cluster is critical. However, companies usually have privacy concerns and hesitate to reveal their technological information. Therefore, we propose a patent recommendation method based on federated learning, which learns a shared recommendation model across companies without direct access to their data and aggregates the preferences of company members in a cluster to identify common technological needs.