| Title |
Aptarnavimo sistemų modeliavimas naudojant fazinio tipo skirstinius / |
| Translation of Title |
Modelling of queuing systems using phase-type distributions. |
| Authors |
Laurikaitytė, Rūta |
| Full Text |
|
| Pages |
45 |
| Keywords [eng] |
Markov models ; phase-type distributions ; moment method ; EM method |
| Abstract [eng] |
Frequently we are unable to use Markov processes to describe real stochastic systems because application streams do not have Poison distribution or the durations of operations do not have exponential distribution. In this study we investigated possibilities of system approximation using mixtures and convolutions of exponential distributions. Approximations of four common distributions (inverse Gaussian, gamma, lognormal and Weibull) have been researched in this project. They were executed with moments and maximum likelihood methods. Accuracy of system approximation improves by increasing number of phases and operations, but then we have an issue of rapidly growing Markov chain difficulty which aggravates the system analysis. 2 and 3 moments matching gives sufficiently good results. Maximum likelihood method suffices with 2 phases and 1000 iterations for reliable results. |
| Dissertation Institution |
Kauno technologijos universitetas. |
| Type |
Master thesis |
| Language |
Lithuanian |
| Publication date |
2015 |
