Dynamic model selection and analysis based on the distribution of power series

In this paper, the traditional likelihood ratio test (LRT) method has been improved. The relative entropy density deviation (REDD) method was introduced for the distribution of power series dynamic models. Through small amounts of data analysis and comparison, the mathematical models can be quickly established based on the collected data. Experiments´ results show that this method can be used to dynamically select the Poisson distribution model, the binomial distribution model, the negative binomial model and so on.