In the continuous cutting processes of a machine tool, the lowest natural frequency, maximum dynamic compliance, average dynamic compliance and Coefficient of Merit of the machine tool would be changed at different degree. Therefore, the research on deterioration trend prediction of the dynamic compliance of a machine tool is very important. However, the relative excitation experiment has much influence on the normal usage of a machine tool, it is difficult to obtain a great deal of experimental data. To solve the problem of small sample size of experiment data, Generalized Hidden Markov Model (GHMM) and gravity method was used to predict the deterioration trend of the dynamic compliance of a machine tool. The research shows that, compared with Hidden Markov Model (HMM), GHMM can deal with the problem of small sample size well. Meanwhile, the precise prediction of all evaluation criterions can decrease the number of relative exciting experiments, and can help to obtain the deterioration trend of the dynamic compliance of a machine tool.