The present invention provides a cross-axis and cross-point modal testing and parameter identification method for predicting the cutting stability, which is used to improve the accuracy of existing prediction methods of cutting stability. The method firstly installs a miniature tri-axial acceleration sensor at the tool tip, and conducts the cross-axis and cross-point experimental modal tests respectively. The measured transfer functions are grouped according to different measuring axes, and the dynamic parameters are separately identified from each group of transfer functions. Then, the contact region between the cutter and workpiece is divided into several cutting layer differentiators. After that, together with other dynamic parameters, all the parameters are assembled into system dynamic parameter matrices matching with the dynamic model. Finally, dynamic parameter matrices including the effects of cross-axis and cross-point model couplings are obtained. Moreover, the acceleration sensor in the method only needs to be installed once.