Evaluation of the efficiency of reference machine learning models for buffer memory prediction when transforming a self-similar input stream of packets into a stream having exponential distribution under the condition of equality of mathematical expectations and median flows
Using machine learning methods, models have been developed to predict the size of the queue depending on the Hurst exponent based on the data obtained when performing the transformation of an input self-similar stream distributed according to the Pareto law into a stream having an exponential distri...
Saved in:
| Main Authors: | , , |
|---|---|
| Format: | Article |
| Language: | Russian |
| Published: |
North-Caucasus Federal University
2023-01-01
|
| Series: | Современная наука и инновации |
| Subjects: | |
| Online Access: | https://msi.elpub.ru/jour/article/view/1358 |
| Tags: |
Add Tag
No Tags, Be the first to tag this record!
|
| Summary: | Using machine learning methods, models have been developed to predict the size of the queue depending on the Hurst exponent based on the data obtained when performing the transformation of an input self-similar stream distributed according to the Pareto law into a stream having an exponential distribution with equal mathematical expectation and equal medians. A comparative analysis of the obtained models is carried out. Each model was examined using the following quality metrics: coefficient of determination, rms regression error, mean absolute error, penalty value, estimated loss. Models that use isotonic regression and support vector methods are the best in terms of the selected quality metrics for methods of transforming the input and output packet streams when the mathematical expectation is equal. For methods of transforming the input and output packet stream with equal medians, linear models are the best. |
|---|---|
| ISSN: | 2307-910X |