Future network condition predictor for network time series data utilizing a hidden Markov model for non-anomalous data and a gaussian mixture model for anomalous data

Patent metadata

Field	Value
Publication number	US-10911318-B2
Application number	US-201615077762-A
Country	US
Kind code	B2
Filing date	Mar 22, 2016
Priority date	Mar 24, 2015
Publication date	Feb 2, 2021
Grant date	Feb 2, 2021

Abstract

System and method embodiments are provided for adaptive anomaly detection based predictor for network data. In an embodiment, a computer-implemented method in a network component for predicting values of future network time series data includes receiving, with one or more receivers, network time series data; determining, with one or more processors, whether an anomaly is detected in the network time series data; generating, with the one or more processors, a prediction associated with the network data according to a primary predictor when no anomaly is detected in the network time series data; generating, with the one or more processors, the prediction associated with the network data according to an alternative predictor when an anomaly in the network time series data is detected; and sending, with one or more transmitters, the prediction to a network controller, wherein the network controller uses the prediction to adjust network parameters.

First claim

What is claimed is:

1. A computer-implemented method in a network component for predicting values of future network time series data, comprising:
receiving, with one or more receivers, first network time series data and second network time series data, both the first network time series data and the second network time series data being historical data associated with traffic patterns over a mobile access network;
detecting, by one or more processors of a single network device, an anomaly in the second network time series data, and no anomaly being detected in the first network time series data;
generating, by the one or more processors of the single network device, a first prediction of a future network condition using the first network time series data according to a Hidden Markov Model (HMM) upon determining that no anomaly is detected in the first network time series data and a second prediction of a future network condition using the second network time series data according to a Gaussian Mixture Model (GMM) upon detecting the anomaly in the second network time series data, the first prediction of the future network condition being generated according to the HMM without relying on the GMM and the second prediction of the future network condition being generated according to the GMM without relying on the HMM; and
sending, with one or more transmitters, the first prediction and the second prediction to a network controller to prompt the network controller to adjust network parameters based on the first prediction and the second prediction.

2. The computer-implemented method of claim 1 , wherein the HMM is trained according to the historical data.

3. The computer-implemented method of claim 1 , wherein the anomaly is detected in the second network time series data based on a likelihood of occurrence of an observed data point.

4. The computer-implemented method of claim 3 , wherein the anomaly is detected in the second network time series data based on a number of calculated likelihood values that fall below a threshold exceeding a predefined number within a specified time period.

5. The computer-implemented method of claim 3 , wherein the likelihood is computed according to the GMM model.

6. The computer-implemented method of claim 5 , wherein the GMM model comprises parameters learned from the historical data.

7. The computer-implemented method of claim 1 , wherein the HMM model comprises parameters learned from the historical data.

8. A network component comprising:
a processor; and
a non-transitory computer readable storage medium storing programming for execution by the processor, the programming including instructions for:
receiving, with one or more receivers, first network time series data and second network time series data, both the first network time series data and the second network time series data being historical data associated with traffic patterns over a mobile access network;
detecting an anomaly in the second network time series data, and no anomaly being detected in the first network time series data;
generating a first prediction of a future network condition using the first network time series data according to a Hidden Markov Model (HMM) upon determining that no anomaly is detected in the first network time series data and a second prediction of a future network condition using the second network time series data according to a Gaussian Mixture Model (GMM) upon detecting the anomaly in the second network time series data, the first prediction of the future network condition being generated according to the HMM without relying on the GMM and the second prediction of the future network condition being generated according to the GMM without relying on the HMM; and
sending, with one or more transmitters, the first prediction and the second prediction to a network controller to prompt the network controller to adjust network parameters based on the first prediction and the second prediction.

9. The network component of claim 8 , wherein the HMM is trained according to historical data.

10. The network component of claim 8 , wherein the anomaly is detected in the second network time series data based on a likelihood of occurrence of an observed data point.

11. The network component of claim 10 , wherein the anomaly is detected in the second network time series data based on a number of calculated likelihood values that fall below a threshold exceeding a predefined number within a specified time period.

12. The network component of claim 10 , wherein the likelihood is computed according to the GMM model.

13. The network component of claim 12 , wherein the GMM model comprises parameters learned from the historical data.

14. The network component of claim 8 , wherein the HMM model comprises parameters learned from the historical data.

15. A non-transitory computer-readable media storing computer instructions for predicting values of future network time series data, that when executed by one or more processors of a single network device, cause the one or more processors to perform the steps of:
receiving, with one or more receivers, first network time series data and second network time series data, both the first network time series data and the second network time series data being historical data associated with traffic patterns over a mobile access network;
detecting an anomaly in the second network time series data, and no anomaly being detected in the first network time series data;
generating a first prediction of a future network condition using the first network time series data according to a Hidden Markov Model (HMM) upon determining that no anomaly is detected in the first network time series data and generating a second prediction of a future network condition using the second network time series data according to a Gaussian Mixture Model (GMM) upon detecting the anomaly in the second network time series data, the first prediction of the future network condition being generated according to the HMM without relying on the GMM and the second prediction of the future network condition being generated according to the GMM without relying on the HMM; and
sending, with one or more transmitters, the first prediction and the second prediction to a network controller to prompt the network controller to adjust network parameters based on the first prediction and the second prediction.

16. The non-transitory computer-readable media of claim 15 , wherein the HMM is trained according to historical data.

17. The non-transitory computer-readable media of claim 15 , the anomaly is detected in the second network time series data based on the GMM model.

18. The non-transitory computer-readable media of claim 15 , wherein the GMM model comprises parameters learned from the historical data.

Assignees

• Futurewei Technologies Inc

Inventors

• Gopalakrishnan Nandu
• Hu Yirui

Classifications

• G06N7/01

  Probabilistic graphical models, e.g. probabilistic networks · CPC title

• H04L41/149

  for prediction of maintenance · CPC title

• H04L41/147

  for predicting network behaviour · CPC title

• H04L41/142

  using statistical or mathematical methods · CPC title

• H04L41/16 Primary

  using machine learning or artificial intelligence · CPC title