Automated intelligent self-organizing network for optimizing network performance

Patent metadata

Field	Value
Publication number	US-10728773-B2
Application number	US-201815880848-A
Country	US
Kind code	B2
Filing date	Jan 26, 2018
Priority date	Jan 26, 2018
Publication date	Jul 28, 2020
Grant date	Jul 28, 2020

Abstract

A Self-Organizing Network (SON) collects data pertaining to a first number of cells of a wireless network. The SON splits the collected data into a second number of groups, and, for each of the second number of groups, repeatedly set a third number of clusters to a different number between a low limit and a high limit. The SON, for each of the settings, clusters the cells into the third number of clusters and trains a deep neural network to perform a regression analysis on the third number of clusters. For each of the second number of groups, the SON also determines an optimum number of clusters based on the regression analyses, re-clusters the cells into the optimum number of clusters; and tunes engineering parameters based on the re-clustering to optimize performance of the wireless network and quality of experience pertaining to the wireless network.

First claim

What is claimed is:

1. A system comprising:
one or more devices configured to:
collect, from nodes of a wireless network, data pertaining to a plurality of cells of the wireless network, wherein the data include at least traffic information for each of the plurality of cells;
divide the collected data into a plurality of groups based on timestamps associated with the collected data, wherein each of the timestamps indicates a time at which the corresponding datum was collected, and wherein each of the groups is associated with a different interval of time;
perform, for each of the groups:
for each integer in a range having a low limit and a high limit, wherein the low limit is greater than one and the high limit is greater than the low limit:
cluster the cells into a number of clusters, wherein the number is equal to the integer; and
train a deep neural network to perform a regression analysis on the number of clusters;

select one of the integers as an optimum number of clusters based on the regression analyses;
re-cluster the cells into the optimum number of clusters; and
modify engineering parameters based on the re-clustering to optimize performance of the wireless network and quality of experience pertaining to the wireless network; and

send a first modified parameter, among the modified parameters, to a network device included in the wireless network, wherein the network device is configured to:

receive the first modified parameter; and
set the received parameter as an operating point of a component of the wireless network.

2. The system of claim 1 , wherein when the one or more devices cluster the cells, the one or more devices apply multiple clustering methods, wherein the multiple clustering methods include at least one of:
K-Means clustering, Fuzzy C-Means clustering, Gaussian Mixed Model, Hierarchical clustering; and Clustering Large Application clustering.

3. The system of claim 1 , wherein when the one or more devices cluster the cells, the one or more devices are further to:
apply the Hungarian Assisted Algorithm Clustering (HAAC).

4. The system of claim 1 , wherein when the one or more devices perform the regression analysis, the one or more devices are to:
determine a difference between an output of the deep neural network and a target value; and
calculate an accuracy value based on the determined difference.

5. The system of claim 4 , wherein when the one or more devices select one of the integers as the optimum number of clusters, the one or more devices are further to:
determine a value of the integer, between the low limit and the high limit, at which the accuracy value is a maximum.

6. A method comprising:
collecting, from nodes of a wireless network, data pertaining to a plurality of cells of the wireless network, wherein the data include at least traffic information for each of the cells;
dividing the collected data into a plurality of groups based on timestamps associated with the collected data, wherein each of the timestamps indicates a time at which the corresponding datum was collected, and wherein each of the groups is associated with a different interval of time;
performing, for each of the groups:
for each integer in a range having a low limit and a high limit, wherein the low limit is greater than one and the high limit is greater than the low limit:
clustering the cells into a number of clusters, wherein the number is equal to the integer; and
training a deep neural network to perform a regression analysis on the number of clusters;

selecting one of the integers as an optimum number of clusters based on the regression analyses;
re-clustering the cells into the optimum number of clusters; and
modifying engineering parameters based on the re-clustering to optimize performance of the wireless network and quality of experience pertaining to the wireless network; and

sending a first modified parameter, among the modified parameters, to a network device included in the wireless network, wherein the network device is configured to:
receive the first modified parameter; and
set the received parameter as an operating point of a component of the wireless network.

7. The method of claim 6 , wherein the data further include one or more of:
network statistics; key performance indicators; or engineering parameters.

8. The method of claim 7 , wherein the network statistics include a number indicating an amount of network traffic of a particular type, wherein the key performance indicators include a Transmission Control Protocol (TCP)-Internet Protocol (IP) session establishment delay, and wherein the engineering parameters include a mechanical or electrical down tilt of an antenna.

9. The method of claim 6 , wherein dividing the collected data into the plurality of groups includes:
dividing the collected data into hourly groups.

10. The method of claim 6 , wherein clustering the cells into the number of clusters includes applying multiple clustering methods, wherein the multiple clustering methods include at least one of:
K-Means clustering, Fuzzy C-Means clustering, Gaussian Mixed Model, Hierarchical clustering; and Clustering Large Application clustering.

11. The method of claim 6 , wherein clustering the cells comprises:
applying the Hungarian Assisted Algorithm Clustering (HAAC).

12. The method of claim 6 , wherein the regression analysis includes:
determining a difference between an output of the deep neural network and a target value; and
calculating an accuracy value based on the determined difference.

13. The method of claim 12 , wherein selecting one of the integers as the optimum number of clusters includes:
determining a value of the integer, between the low limit and the high limit, at which the accuracy value is a maximum.

14. The method of claim 6 , wherein re-clustering the cells includes:
applying the Hungarian Assisted Algorithm Clustering (HAAC) to different labels pertaining to different intervals at which the data are collected.

15. The method of claim 6 , wherein modifying the engineering parameters includes:
determining values, of the engineering parameters, that are closest to target values, wherein the target values are associated with key performance indicators.

16. Non-transitory computer-readable media comprising computer-executable instructions, when executed by one or more processors, cause the one or more processors to:
collect, from nodes of a wireless network, data pertaining to a plurality of cells of the wireless network, wherein the data include at least traffic information for each of the cells;
divide the collected data into a plurality of groups based on timestamps associated with the collected data, wherein each of the timestamps indicates a time at which the corresponding datum was collected, and wherein each of the groups is associated with a different interval of time;
perform, for each of the groups:
for each integer in a range having a low limit and a high limit, wherein the low limit is greater than one and the high limit is greater than the low limit:
cluster the cells into a number of clusters, wherein the number is equal to the integer; and
train a deep neural network to perform a regression analysis on the third number of clusters;

select one of the integers as an optimum number of clusters based on the regression analyses;
re-cluster the cells into the optimum number of clusters; and
modify engineering parameters based on the re-clustering to optimize performance of the wireless network and quality of experience pertaining to the wireless network; an

Assignees

• Verizon Patent & Licensing Inc

Inventors

• Ouyang Ye
• Su Le
• Iyer Krishna Pichumani
• Schmidt Christopher M

Classifications

• G06N3/045

  Combinations of networks · CPC title

• G06N3/0499

  Feedforward networks · CPC title

• G06N3/09

  Supervised learning · CPC title

• H04W24/08

  Testing, {supervising or monitoring} using real traffic · CPC title

• H04W24/02 Primary

  Arrangements for optimising operational condition · CPC title