2017 The 5th International Symposium on Project Management (ISPM 2017)

117 / 2017-06-18 18:47:08

Parameter Estimation of Discrete-continuous Mixture Distributions

EM algorithm,mixture distribution,parameter estimation

The paper deals with the problem of discrete-continuous mixture distribution parameter estimation. Using the principle of maximal likelihood estimation, the likelihood function of unknown parameters is given, the Q function and the updating formulas of the unknown parameters are deduced when using the EM algorithm for parameter estimation, and the flowchart of EM algorithm is given. Then, the EM algorithm is used on numerical simulation experiments, and the validity of EM algorithm in solving these problems is verified. At the same time, this paper finds that the estimation accuracy of EM algorithm was very affected by the initial value.