Communications Network Configuration via Probabilistic KPI Graphs
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
The challenge in configuring communications networks, particularly cellular networks, is identifying relevant features and configuration changes to improve specific Key Performance Indicators (KPIs, such as VoLTE call drop rates and streaming media quality) amidst a vast number of daily network adjustments, which becomes more complex with the deployment of 5G and further cell densification.
Innovation Solution
A method involving the construction of a probabilistic graph from operational data of network nodes, defining actions and states, and determining transitions between these states to identify actions that lead to desired KPI improvements, using historical data to infer influential features and configuration changes.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If multiple ML models are used to analyze KPIs and test different feature combinations, then the ability to understand critical features and determine improvement combinations improves, but the complexity of the system increases and determining suitable feature combinations becomes more challenging
Solution Approach 1:
The patent segments the complex task of KPI improvement into distinct components: an exploration module that discovers relevant features through probabilistic graphs, and an exploitation module that optimizes configurations based on discovered features. This segmentation divides the monolithic ML approach into manageable modules, reducing overall system complexity while maintaining identification accuracy.
Solution Approach 2:
The patent introduces probabilistic graphs as an intermediary representation that bridges operational data and KPI analysis. Instead of directly using multiple complex ML models on raw data, the system first constructs probabilistic graphs that capture feature relationships, then performs analysis on this simplified intermediate structure. This intermediary layer reduces the complexity of subsequent analysis tasks.
2Productivity
If technicians manually monitor and reconfigure network parameters daily, then general network performance metrics such as coverage and inter-cell interference can be improved, but the ability to identify relevant features and configuration changes for specific KPIs deteriorates due to the vast number of adjustments
Solution Approach 1:
The patent implements feedback mechanisms where the system continuously monitors KPI performance and uses this feedback to guide further exploration and exploitation. The exploration module uses feedback from KPI changes to update probabilistic graphs and identify new relevant features, while the exploitation module uses feedback to refine configuration recommendations. This feedback loop enables automated adaptation without manual intervention.
Solution Approach 2:
The system performs self-service by automatically discovering relevant features and determining optimal configurations without requiring technician expertise in identifying specific KPI-driving factors. The probabilistic graph construction and analysis are performed autonomously, allowing the system to serve itself in identifying and implementing improvements while technicians focus on higher-level decisions.
3Adaptability or versatility
If the network undergoes further cell densification for 5G deployment, then network capacity and coverage improve, but the complexity of identifying relevant configuration changes increases beyond manual manageability
Solution Approach 1:
The patent applies dynamics by making the system adaptable to changing network conditions through continuous probabilistic graph updates. As the network evolves with 5G cell densification, the system dynamically reconstructs probabilistic graphs from current operational data, automatically adapting to new network topologies and configurations. This dynamic approach allows the system to scale with network complexity without requiring proportional increases in manual management effort.
Data Source
AI summary
A method of configuration in a communications network includes: obtaining operational data relating to operation of at least one network node, the operational data including, for each of a plurality of time periods, data relating to at least one configuration parameter of the network node and data relating to at least one performance indicator of the network node; defining a plurality of actions, wherein each action corresponds to a specific set of values of the at least one configuration parameter in two successive time periods; and defining a plurality of states, wherein each state corresponds to one specific set of values of the operational data. A probabilistic graph is constructed from the operational data, the probabilistic graph indicating, for a plurality of the defined states, a probability that an action will lead to a transition to a respective other state.