US 11,888,703 B1
Machine learning algorithms for quality of service assurance in network traffic
Bernardo Huberman, Palo Alto, CA (US); and Scott H. Clearwater, Menlo Park, CA (US)
Assigned to Cable Television Laboratories, Inc., Louisville, CO (US)
Filed by CABLE TELEVISION LABORATORIES, INC., Louisville, CO (US)
Filed on Feb. 16, 2021, as Appl. No. 17/176,654.
Application 17/176,654 is a continuation of application No. 16/569,021, filed on Sep. 12, 2019, granted, now 10,931,538.
Claims priority of provisional application 62/856,615, filed on Jun. 3, 2019.
Claims priority of provisional application 62/730,966, filed on Sep. 13, 2018.
This patent is subject to a terminal disclaimer.
Int. Cl. H04L 41/147 (2022.01); H04L 43/0882 (2022.01); H04L 43/067 (2022.01); H04L 41/16 (2022.01); G06F 17/18 (2006.01); H04L 43/0823 (2022.01); H04L 43/106 (2022.01); H04L 41/142 (2022.01)
CPC H04L 41/147 (2013.01) [G06F 17/18 (2013.01); H04L 41/142 (2013.01); H04L 41/16 (2013.01); H04L 43/067 (2013.01); H04L 43/0847 (2013.01); H04L 43/0882 (2013.01); H04L 43/106 (2013.01)] 26 Claims
OG exemplary drawing
 
1. A method for predicting future idle capacity of a network comprising:
measuring traffic volumes on a network during multiple different time windows to obtain a plurality of first data sets, each of the first data sets having associated with it a future traffic rate;
mathematically generating a plurality of vectors, each of the vectors corresponding to and having a higher dimensionality than one of the first data sets;
encoding each of the vectors, thereby generating a plurality of encodings corresponding to the plurality of first data sets;
determining a future traffic value based on the future traffic rate of each of the first data sets that generated a specific encoding selected from the plurality of encodings;
associating the future traffic value with the specific encoding and storing in memory;
repeating the steps of determining and associating for each encoding of the plurality of encodings, thereby generating a plurality of encoding-to-future traffic value associations; measuring current traffic volumes on the network to produce a current traffic volume data set;
creating a current traffic encoding representing the current traffic volumes within the current traffic volume data set;
matching the current traffic encoding to one or more of the encoding-to-future traffic value associations to obtain the future traffic value, which is a predicted future traffic rate for the current traffic volume data set;
subtracting the predicted future traffic rate from a maximum network capacity to determine a future idle capacity; and
reallocating at least a portion of the future idle capacity of the network.