OBJECTIVE： To provide technical and practical basis for data-driven pharmaceutical management model. METHODS： The information of drug attribute classification and the way of network acquisition were studied， and the construction of drug attribute classification knowledge base was realized by using network crawler technology. On this basis， by using NLP technology and regular expression matching technology， the information of the dictionary table of medicines in HIS and the drug name， specifications and manufacturers in the drug attribute classification knowledge base were compared one by one. The correspondence of drug codes between “HIS database” and “drug attribute classification knowledge base” was realized. Finally， drug attribute classification knowledge base can be used to practice drug use analysis in medical institutions. RESULTS： Established drug attribute classification knowledge base can rapidly and accurately analyze the use of antibiotics， evaluate the reasonability of drug procurement and screen unreasonable drug use case rapidly. “Medical Institution Pharmaceutical Management Analysis and Decision-Making System” was established by using knowledge base of drug attribute classification and control technology. All prescription reviews had been realized with this system， and the main problems of irrational drug use in clinic， such as over-dose， over-course of treatment， combined use of main drugs with the same function and over-indication drug use could be solved. Through the practice and application of the prescription comment model， the above problems in our hospital’s prescription comment were almost nonexistent. Antibacterial use indicators were also in line with national requirements. CONCLUSIONS： It is beneficial to the information construction of pharmacy administration by linking actual use of drugs in medical institutions with drug-related knowledge base through modern information technology and taking the actual use of drugs in medical institutions as the object of analysis. In the future， it is necessary to build knowledge bases for different purposes and types based on the requirements of pharmaceutical administration. It lays a foundation for the practice of data-driven pharmaceutical administration.