Software fault intelligent classification method and readable storage medium

The invention provides a software fault intelligent classification method, which comprises: enabling preprocessing to adopt an unsupervised clustering method, and generating classification as required; reconstructing keywords by adopting a means of replacing synonyms, synonyms and anonyms, and constructing a new topic center by adopting clustering center keywords; and adopting an optimized k-means model to create a subclass number window function. The invention further provides a readable storage medium. The method has the advantages that in classification of new problems caused by new functions of the intelligent automobile market, the problem types do not need to be preset, unexpected software fault types can be effectively explored, categories can be classified according to the on-demand automobile lexicon, and the return rate is reduced by about 12%; keywords are reconstructed by adopting a few-to-many means, a topic center thought can be effectively highlighted, new topics are constructed by clustering h