Research On Mixed Discrete-continuous Model Based On Vine Copula

With the development of society,the size and dimensions of data are increasing rapidly.The increase in data size imposes higher requirements on the computational intensity and time of analysis.Moreover,there are very few cases of pure discrete variables or pure continuous variables in multivariate data.Therefore,the problem of multivariate mixed discretecontinuous data has become a hot topic in current research.At present,the model based on copula can be specifically adapted to solve this kind of problems.When it comes to mixed discrete-continuous problems,it is difficult to choose a suitable copula function,and the increase in dimension also greatly increases the calculation cost.This paper mainly studies the multivariate mixed discrete-continuous model based on vine copula.Aiming at the air quality problem in Suzhou,three five-dimension mixed discrete-continuous models were established and analyzed.This paper introduces the theory of copula,and combines the normal copula with vinecopula structure to get the vine-copula model with pair copula fixed.This is mainly due to the adaptability of normal copula in the case of binary continuous variables,binary discrete variables,binary mixed discrete-continuous variables,and the flexibility of vine-copula structure in the decomposition of multiple variables.Finally,this paper adopts copula-based model construction method to analyze the multivariate AQI data of Suzhou,including two discrete variables and three continuous variables.Three models were established in this paper,namely normal copula model,vine-copula model based on D vine and vine-copula model based on R vine.In terms of performance of the three models,the two vine-copula models performed better,and the vine copula based on D vine model explained the best.The calculation cost of the normal copula model is slightly higher,but it is also a well-fitted parameter model.