The invention discloses a knowledge graph construction method based on an improved BERT model, and the method comprises the steps: obtaining multi-source heterogeneous data, and storing the obtained multi-source heterogeneous data to a big data platform; carrying out knowledge modeling on the acquired multi-source heterogeneous data based on the ontology; acquiring a knowledge on the basis of a BERT-BiGRU-CRF model; performing entity fusion and attribute fusion on the obtained heterogeneous data sources based on knowledge fusion of word vectors; and performing knowledge storage by adopting a graphic database. The enterprise knowledge graph is adopted to break through a traditional calculation mode, internal and external information and data in the field of enterprise operation management are deeply integrated, early warning potential risks can be more effectively mined in the later period, enterprise operation management efficiency is improved, and improvement of accuracy of subsequententerprise operation management risk intelligent diagnosis is facilitated.