SS&IAGA-EM-based Algorithm for Fitting a Continuous PH Distribution

It is an important and difficult task in model analysis of traffic engineering to fit with the general distribution or test data whether which fit the distribution or not via PH distribution. Although there are lots of methods to fit continuous PH distribution, which are all lack of efficiency and numerical stability. For giving consideration to both fitting effect and efficiency of a continuous PH distribution, we apply the maximum likelihood method to the dense subset HErD of PH distribution and design a SS & IAGA-EM algorithm for study. In the data fitting test for long-tailed distribution function, partial peak distribution function and heavy-tailed distribution function with a sample size of 104, the maximum error of the algorithm is 7.32%. When operated in a standard PC with 2.5 GHz Pentium CPU running under the operating system of Windows XP, the maximum operating time of the algorithm is 100 s, which meets the demand for effectiveness and efficiency.