The idea of classification automation is the inevitable result of the continuous progress of information processing technology. Library classification system is a system to group the collected record groups, which can make related topics belong to the same system. Clearly dividing groups or classes can reduce the number of items that users browse. Computers themselves can't understand ideas, but the ideas in files can be represented by words and numbers, which can be processed and compared by computers, and then files can be classified according to the similarities and differences of words. The research of automatic classification includes two aspects: ① Derive the method of automatic classification system, that is, first express each file as a directory-word glossary, and then compare each file with every other file collected according to its * * * words to get subgroups; Each grouping is a category in the classification scheme. ② The mathematical method for calculating the probability of document classification is studied in detail, that is, the automation of classification.