# US 2019/0245754 A1 · Classification and Relationship Correlation Learning Engine for the Automated Management of Complex and Distributed Networks > Agents: this data is available as JSON from the patentgrep API and MCP server — see https://patentgrep.com/llms.txt - Number: US 2019/0245754 A1 - Document ID: US20190245754A1 - Status: Published application - Kind: A1 - Application number: 16/214,421 - Filed: 2018-12-10 - Published: 2019-08-08 - Earliest claimed: 2018-02-05 - Assignees: Crenacrans Consulting Services - Applicants: Crenacrans Consulting Services - Inventors: James W. Greene, JR. - Web page: https://patentgrep.com/patent/US20190245754A1 ## CPC - H04L 41/16: using machine learning or artificial intelligence - G06N 3/084: Backpropagation, e.g. using gradient descent - G06N 20/00: Machine learning - H04L 41/142: using statistical or mathematical methods - H04L 41/145: {involving simulating, designing, planning or modelling of a network} - G06N 3/006: based on simulated virtual individual or collective life forms, e.g. social simulations or particle swarm optimisation [PSO] - G06N 5/01: Dynamic search techniques; Heuristics; Dynamic trees; Branch-and-bound - H04L 67/10: in which an application is distributed across nodes in the network (software deployment G06F8/60; multiprogramming arrangements G06F9/46) - H04L: TRANSMISSION OF DIGITAL INFORMATION, e.g. TELEGRAPHIC COMMUNICATION (arrangements common to telegraphic and telephonic communication H04M) - G06N: COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS ## Abstract The present disclosure describes a method, system, and apparatus for using a machine learning system to configure and optimize complex, distributed computer networks. The machine learning system receives an input related to a computer network and classifies the input using either a supervised learning approach or an unsupervised learning approach. From the classification of the input, the machine learning system builds a first training domain and determines a steady state network configuration for the computer network. After determining a steady state network configuration for the computer network, the machine learning system receives a plurality of inputs from one or more sensors or agents distributed throughout the computer network. The machine learning system compares the plurality of inputs to the steady state network configuration to detect a deviation from the first steady state network configuration. When a deviation from the steady state network configuration is detected, the machine learning system remediates the problem to return the computer network to the steady state network configuration. ## Claims ### Claim 1 A method for managing a first network, the method comprising: obtaining, at a first server, one or more features from the first network; classifying and labelling the one or more features obtained from the first network; defining, by the first server, a first training domain using the classified and labelled one or more features; creating, by the first server, a first steady state model of the first network using, in part, the first training domain; monitoring, by the first server, the first network by comparing a first current state model to the first steady state model; and remediating, by the first server, the first network when the first current state model of the first network deviates from the first steady state model. ### Claim 2 (depends on 1) The method of claim 1, wherein obtaining one or more features from the first network further comprises: transmitting, from the first server, a first request for information to one or more first devices; receiving, at the first server, a first response to the first request for information from one or more first devices; analyzing, by the first server, first network traffic between at least one second device and one third device; determining, by the first server, one or more features of the first network using at least one of the first response and the analysis of the first network traffic; and creating, by the first server, a first configuration of the first network based, in part, on the one or more features. ### Claim 3 (depends on 2) The method of claim 2, wherein analyzing the first network traffic further comprises: inspecting, by the first server, one or more packets exchanged between the second device and the third device; determining, by the first server, a classification for the first network traffic; and monitoring, by the first server, whether the classification for the first network traffic is permitted to traverse the first network between the second device and third device. ### Claim 4 (depends on 2) The method of claim 2, further comprising: updating, by the first server, the first configuration of the first network based on further monitoring of the first network. ### Claim 5 (depends on 1) The method of claim 1, wherein monitoring the first network further comprises: comparing the first current state model of the first network to the first steady state model to determine whether the first network comports to baseline configuration parameters; identifying one or more training domains that are non-compliant when the first network does not comport to baseline configuration parameters; identifying one or more parameters within the one or more non-compliant training domains that are non-compliant; providing one or more instructions to remediate the non-compliant parameters. ### Claim 6 (depends on 5) The method of claim 5, wherein the first current state model of the first network is defined using one or more optimal machine learning algorithms. ### Claim 7 (depends on 5) The method of claim 5, wherein the one or more non-compliant training domains are identified using neural network back propagation. ### Claim 8 (depends on 5) The method of claim 5, wherein the one or more non-compliant parameters are identified using neural network back propagation. ### Claim 9 (depends on 5) The method of claim 5, further comprising: determining whether the first network has returned to the first steady state model in response to the one or more instructions to remediate the non-compliant parameters by comparing a second current state model to the first steady state model. ### Claim 10 (depends on 5) The method of claim 5, further comprising: determining additional remedial action to correct the one or more non-compliant parameters when the first network has not returned to the first steady state model in response to the one or more instructions. ### Claim 11 (depends on 1) The method of claim 1, wherein classifying and labelling the one or more features obtained from the first network further comprises: determining, by the first server, whether the one or more features associated with the first network have been classified; when the one or more features associated with the first network have been classified, determining whether the one or more features have been labelled; and in response to determining that the one or more features have been labelled, storing the one or more labelled features in a memory. ### Claim 12 (depends on 11) The method of claim 11, further comprising: determining an appropriate label for the one or more features when the one or more features have not been labelled; and applying the appropriate label to the one or more features. ### Claim 13 (depends on 11) The method of claim 11 further comprising: determining an appropriate classification for the one or more features when the one or more features have not been classified; and applying the appropriate classification to the one or more features. ### Claim 14 (depends on 13) The method of claim 13, wherein determining an appropriate classification for the one or more features further comprises: determining, by the first server, an optimal operating range for the one or more features; querying, by the first server, the one or more devices with respect to the one or more features to obtain additional information; receiving, at the first server, additional information from the one or more devices about the one or more features; storing, at the first server, the additional information received from the one or more devices; applying a classification to the one or more features based, in part, on the additional information. ### Claim 15 (depends on 1) The method of claim 1, wherein defining the first training domain further comprises: extracting a first set of features and a second set of features from the one or more features obtained from the first network; determining whether a relationship exists between the first set of features and the second set of features by applying first training data; determining that the first set of features and the second set of features are related if a behavior of at least one of the second set of features changes in response to at least one of the first set of features being modified. ### Claim 16 (depends on 15) The method of claim 15, further comprising: assigning the first set of features and the second set of features to the first training domain when the first set of features and the second set of features are related. ### Claim 17 (depends on 1) The method of claim 1, wherein creating the first steady state model further comprises: determining whether a relationship exists between the first training domain and a second training domain; combining the first training domain and the second training domain into a superset domain when a relationship exists between the first training domain and the second training domain; determining an optimal machine learning algorithm for a plurality of training domains and the superset domain; and defining, by the first server, the first steady state model of the first network using at least one of the plurality of training domains and the superset domain. ### Claim 18 A method for configuring and optimizing a computer network, the method comprising: receiving, at a first server, at least one input variable related to the computer network, wherein the at least one input variable is undefined; classifying, by the first server, the at least one input variable using at least one of a supervised learning approach and an unsupervised learning approach; building, by the first server, a first training domain using the at least one classified input variable; determining, by the first server, a first steady state network configuration for the computer network; receiving, at the first server, a plurality of second inputs from one or more sensors monitoring the computer network; comparing, by the first server, the plurality of second inputs to the first steady state network configuration to detect a deviation from the first steady state network configuration; and remediating, by the first server, the computer network to return to the first steady state network configuration when a deviation from the first steady state network configuration is detected. ### Claim 19 (depends on 18) The method of claim 18, further comprising: classifying, by the first server, the plurality of second inputs to allow identification of relevant features; labelling, by the first server, the plurality of second inputs to allow identification of relevant features; and defining, by the first server, the plurality of second inputs maximum thresholds to allow identification of relevant features. ### Claim 20 (depends on 18) The method of claim 18, further comprising: creating, by the first server, one or more training domains from common groups of related data; creating an optimal mathematical model by applying one or more machine learning algorithms to the one or more training domains. ### Claim 21 (depends on 20) The method of claim 20, further comprising: creating an optimal mathematical model for supersets of training domains using machine learning methods; creating a single mathematical model from the one or more training domains using machine learning methods. ### Claim 22 (depends on 18) The method of claim 18, further comprising: determining a steady-state model for a training system using machine learning techniques; learning behavior in real-time using machine learning techniques to predict future state models. ### Claim 23 (depends on 18) The method of claim 18, further comprising: providing, by the first server, a notification to a first user of the deviation from the first steady state network configuration. ### Claim 24 (depends on 18) The method of claim 18, further comprising: maintaining maximum and minimum thresholds for a steady state model using statistical control processes. ### Claim 25 (depends on 18) The method of claim 18, further comprising: comparing, by the first server, real-time and steady state models through statistical control processes. identifying, by the first server, one or more non-compliant training domains using neural networks; identifying, by the first server, non-compliant features within one or more training domains using neural networks. ### Claim 26 (depends on 18) The method of claim 18, further comprising: providing, by the first server, to a plurality of second systems, feedback related to non-compliance with steady-state; receiving, by the first server, from a plurality of second systems, response to feedback related to compliance with steady-state. ### Claim 27 (depends on 18) The method of claim 18, further comprising: presenting data input, manipulated and stored by the first server and a plurality of second systems in storage mediums via mobile or graphical representations of data.