Abstract: Intelligent decision-making is the core research direction in the field of intelligent ventilation in mines,and the prerequisite for achieving intelligent decision-making is effective processing,analysis,and representation learning of data.In re sponse to the issues of low data utilization and insufficient knowledge reasoning capability in mine intelligent ventilation sys tems,this study constructs a knowledge ontology framework for mine ventilation based on the Coal Mine Safety Regulations (2022 Edition).It systematically evaluates the performance characteristics of knowledge representation learning models such as TransE,TransR,RESCAL,DistMult,and ComplEx,achieving vector representation and deep information mining of ventilation knowledge.By using loss functions and evaluation metrics,different dimensional knowledge representation methods were as sessed,determining that the TransE model has the optimal applicability in mine ventilation knowledge representation.Experi mental results show that the TransE model outperforms other models in terms of representation learning efficiency and accura cy,exhibiting a lower average loss value during the representation learning process,with similar ventilation entities showing sig nificant clustering characteristics in the vector space,and optimal performance when the embedding vector is 40 dimensions. This model effectively represents ventilation knowledge,providing high-precision knowledge reasoning support for intelligent de cision-making systems.

Key words: mine ventilation,deep learning,machine training,knowledge representation,knowledge graph