Machine learning systems and methods for computer modeling and optimization of telecommunications networks
Machine learning systems and methods for computer modeling and optimization of telecommunications networks are provided. The system processes information relating to a telecommunications network, such as a cellular telephone network, and generates a directed network graph that represents the network as a series of connected edges and nodes with associated attributes such as transmitter power, antenna tilt, radio resource control information, handover attempts, collisions, and other parameters. The network graph is stored in a graph database, and is processed by one or more network graph model inference modules which apply one or more specially-trained artificial intelligence (AI) models to the network graph to generate an optimized version of the network graph that can be utilized to improve the telecommunications network. A modeling visualization interface is also provided, and allows users of the system to visualize modeling and optimization performed by the system.
1 . A machine learning system for computer modeling and optimization of a telecommunication network, comprising:
an artificial intelligence (AI) modeling processor in communication with a telecommunication network, the AI modeling processor:
receiving information relating to operation of a cluster of communication nodes within the telecommunication network;
processing the information to generate a directed network graph that represents the telecommunication network as a series of connected edges and nodes;
storing the directed network graph in a graph database;
processing the directed network graph using an AI model to generate an optimized version of the network graph, the AI model trained to optimize at least one attribute relating to the telecommunication network; and
generating and displaying a visualization of the optimized version of the network graph in a visualization interface.
2 . The system of claim 1 , wherein the telecommunication network comprises at least one of a wireless network, a cellular telephone network, or a radio access network (RAN).
3 . The system of claim 2 , wherein each node of the directed network graph represents a cell within the cellular telephone network and each edge of the directed network graph represents an interaction between two cells of the cellular telephone network.
4 . The system of claim 3 , wherein each node of the directed network graph stores geographical data representing a physical location and a coverage area of an associated cell, configuration data representing an operational parameter of the cell, and performance data representing a performance metric of the cell.
5 . The system of claim 4 , wherein the geographical data comprises one or more of geographical coordinates, sector azimuth, or a center of radiation.
6 . The system of claim 4 , wherein the configuration data comprises one or more of power settings, antenna tilt data, frequency band data, physical cell identity, or a licensed feature setting.
7 . The system of claim 4 , wherein the performance data comprises one or more of traffic load data, physical resource block utilization data, signal quality data, throughput data, or voice experience data.
8 . The system of claim 1 , wherein the AI model comprises at least one of a graph neural network (GNN), a graph attention (GAT) network, a message passing neural network (MPNN), a convolutional neural network (CNN), a variational autoencoder (VAE), a conditional VAE (CVAU), a multi-layer perceptron (MLP), or a transformer network.
9 . The system of claim 1 , wherein the at least one attribute comprises transmitter power, antenna tilt, radio resource control information, handover attempts, or collisions.
10 . The system of claim 1 , wherein the AI modeling processor executes a network to graph extraction-transform-load (ETL) pipeline to populate the graph database using raw telecommunication data from the telecommunication network.
11 . The system of claim 1 , wherein the AI modeling processor trains the AI model in a training module and deploys a validated AI model in an inference module, wherein the training module is isolated from the inference module.
12 . The system of claim 1 , further comprising a radio access network (RAN) application programming interface (API) executed by the AI modeling processor for obtaining live status information from a RAN.
13 . The system of claim 12 , further comprising a client API gateway executed by the AI modeling processor for interfacing with a computer system operated by a customer engineering team.
14 . A machine learning method for computer modeling and optimization of a telecommunication network, comprising:
receiving by an artificial intelligence (AI) modeling processor information from a telecommunication network relating to operation of a cluster of communication nodes within the telecommunication network;
processing the information to generate a directed network graph that represents the telecommunication network as a series of connected edges and nodes;
storing the directed network graph in a graph database;
processing the directed network graph using an AI model to generate an optimized version of the network graph, the AI model trained to optimize at least one attribute relating to the telecommunication network; and
generating and displaying a visualization of the optimized version of the network graph in a visualization interface.
15 . The method of claim 14 , wherein the telecommunication network comprises at least one of a wireless network, a cellular telephone network, or a radio access network (RAN).
16 . The method of claim 15 , wherein each node of the directed network graph represents a cell within the cellular telephone network and each edge of the directed network graph represents an interaction between two cells of the cellular telephone network.
17 . The method of claim 16 , wherein each node of the directed network graph stores geographical data representing a physical location and a coverage area of an associated cell, configuration data representing an operational parameter of the cell, and performance data representing a performance metric of the cell.
18 . The method of claim 17 , wherein the geographical data comprises one or more of geographical coordinates, sector azimuth, or a center of radiation.
19 . The method of claim 17 , wherein the configuration data comprises one or more of power settings, antenna tilt data, frequency band data, physical cell identity, or a licensed feature setting.
20 . The method of claim 17 , wherein the performance data comprises one or more of traffic load data, physical resource block utilization data, signal quality data, throughput data, or voice experience data.
21 . The method of claim 14 , wherein the AI model comprises at least one of a graph neural network (GNN), a graph attention (GAT) network, a message passing neural network (MPNN), a convolutional neural network (CNN), a variational autoencoder (VAE), a conditional VAE (CVAU), a multi-layer perceptron (MLP), or a transformer network.
22 . The method of claim 14 , wherein the at least one attribute comprises transmitter power, antenna tilt, radio resource control information, handover attempts, or collisions.
23 . The method of claim 14 , further comprising executing, by the AI modeling processor, a network to graph extraction-transform-load (ETL) pipeline to populate the graph database using raw telecommunication data from the telecommunication network.
24 . The method of claim 14 , further comprising training, by the AI modeling processor, the AI model in a training module and deploying a validated AI model in an inference module, wherein the training module is isolated from the inference module.
25 . The method of claim 14 , further comprising obtaining live status information from a radio access network (RAN) using a RAN application programming interface (API).
26 . The method of claim 25 , further comprising interfacing the AI modeling processor with a computer system operated by a customer engineering team using a client API.