In order to guide the fatigue test of small samples and find the best fitting method of S-N curve under small samples, an improved classical sample information aggregation method is proposed. Based on the consistency of the fatigue life probability quantiles of the specimens under different stress levels, data sharing and fusion methods are adopted to realize the application of sample data information aggregation under different stress levels. According to the linear relationship between stress and fatigue life, the improved sample information aggregation method is used to parameterize and gradually search the average fatigue life under each stress level in the small sample data to obtain the optimal value of fatigue life under different stress levels. The least squares method is used to fit the S-N curve. The fatigue characteristics of the S-N curve are compared and analyzed with different stress levels as the benchmark. The comparison and analysis results show that the maximum relative error of the curves fitted with the improved method and the traditional group method is less than 5%, and the range of predicted fatigue life error is the smallest, which shows that the improved method promotes the reliability of fatigue analysis of small samples.