Applying case-based reasoning in high-speed machining can utilize previous cases and experience in machining a new work piece. This is helpful in overcoming the shortage of available high-speed machining data, and has significance for extending the application of HSM technology. For case representation, a model of case in HSM with relational data mode is proposed in the paper. With analysis of the characteristics of attributes and domain knowledge of HSM, the local similarities of case attributes are sorted into four classes: the numerical type, the enumerative type, the independent type, and the dependent type. The global similarity of cases is calculated by a composite similarity measure.