Analysis of Journal Impact Factors Based on Generalized Extreme Value Model

In order to achieve analysis and prediction of journal impact factors, we take Nature magazine as an example in this paper. Firstly, we analyze the rationality of model selection by constructing a Generic Extreme Value model. Next, we calculate the return levels under different return periods and compare them with the data from the test years, to analyze the predictive effect of the Generic Extreme Value model on impact factors. Finally, we analyze the stability of the Generic Extreme Value model in different starting years. The results are shown that: (1) The Generic Extreme Value model and the Gumbel Distribution model are feasible to impact factors. The Generic Extreme Value model does not restrict on the shape parameters and has good flexibility. (2) The return levels and the confidence interval of the return levels can predict the impact factors, and as the return periods increase, the accuracy of the prediction gradually decreases. (3) The Generic Extreme Value model is not sensitive to the starting years, and has good stability.