1. College of Earth Sciences，Jilin University，Changchun 130061，China; 2. Jilin Seismic Station，Jilin Earthquake Agency，Changchun 130117，China; 3. Helong Seismic Station，Jilin Earthquake Agency，Changchun 130200，China; 4. Siping Branch，Jishi Media Co． Ltd．，Siping 136000，Jilin China; 5. Information Center，Jilin Earthquake Agency，Changchun 130117，China

In order to verify the fitting effect of singular spectrum analysis in geomagnetic data，the authors conduct singular spectrum analysis by using Z，H，D and F components，which have been collected by the FHDZ--M15 instrument located in Changchun Geomagnetic Station during the period from January 2017 to July 2019． By constructing trajectory matrix，decomposing singular values and reconstructing series，we extract the trend term，noise term and finally obtain the fitting values of original series． We create unary linear regression as a contrast model and calculate the fitting values． ＲMSE ( root mean square error) ，MAE ( mean absolute error) and MSPE ( meansquare percentage error) have been used to make error analysis on the fitting results of singular spectrum analysis and unary linear regression analysis． The results show that the fitting results of singular spectrum analysis have higher accuracy． The AＲMA model has been constructed to make value prediction on the singular spectrum fitting sequence． The confidence interval has been given and the prediction performance has been analyzed． AＲMA prediction model shows that changes in the trend of the predicted value is consistent with that of the original sequence，and the predicted value does not exceed the confidence interval，which indicates that there will be no geomagnetic anomaly in the prediction period．