Solar flare is an important space weather event, and prediction for the parameters of it is important to the warning of Sudden Ionospheric Disturbance (SID). It is well known that the revolution of solar soft X-ray flares follows some pattern, especially the long duration ones. Through events analysis, it is found that the early stage of rising phase of the solar flare always indicates the peak flux. After checking the process of recent events, the pattern of flux of flare developing become clear that the rising rate flux decrease continuously until reaching the peak. To simulate this process, a mixed fitting model was built and the basic idea stands that the time X-ray data is fed into the model and relative prediction result could be derived from each group of data. When the difference between adjacent two is within point scale, the fitting parameter is then determined and stored in the model. The prediction of the flare will be made with these parameters. The evaluation of the model was done with the big flare in 23rd solar cycle and the result is acceptable with the error less than 20% and the peak time could also be well predicted. The decreasing phase of the flare was also investigated and the result showed that most of the flares go down following a minus e-fold pattern. Although the appearance of each flare is different, its decreasing curve could be simulated with the data after peak time. The fitting parameters would be derived through 10 data points. The result of the fitting model indicates that the ending time of flare could be well predicted. Judging from the forecast results with historic data, the prediction method has a certain validity and practicability. The dynamic fitting model was also tested with newly occurred flare. The size and peak time were well predicted, which proved that this method could be applied to prediction practically.