Telecommunication Network Event Data Management

Resolve Bottlenecks,
Find Innovative Solutions
Generate Solutions

Solution Overview

Problem

The complexity of managing network event data across different vendors in telecommunication networks, where existing Machine Learning applications struggle to consume data due to vendor-specific metadata differences, leading to high costs and time requirements for developing specialized applications.

Innovation Solution

A method that identifies common features across vendors by generating feature description vectors, constructing graphs, and partitioning them into clusters to normalize network event features, allowing a single shared Machine Learning application to make recommendations across multiple vendor products.

Engineering Contradictions & Design Principles

VSEngineering Contradiction Analysis

1Measurement precision

If multiple specialized Machine Learning applications are developed for different vendors, then data processing accuracy for each vendor is improved, but development cost and time increase significantly

Engineering Contradiction:
Improvedata processing accuracyVSAvoiddevelopment time
Core Design Contradiction:
Measurement precisionVSLoss of time

Solution Approach 1:

The patent implements a universal Machine Learning application that can process data from multiple vendors by introducing a metadata-based feature derivation layer. This layer automatically adapts vendor-specific data formats and metadata schemas to a standardized internal representation, allowing a single ML model to serve multiple vendors without requiring separate specialized applications for each vendor.

Inventive Principle:
Principle #6Universality (Multi-functionality)

Solution Approach 2:

The patent introduces a metadata intermediary layer that sits between vendor-specific data sources and the Machine Learning application. This metadata layer acts as a mediator that translates and harmonizes different vendor data formats, column names, and structures into a unified feature set that the ML application can consume, eliminating the need for vendor-specific processing pipelines.

Inventive Principle:
Principle #24Intermediary (Mediator)

2Adaptability or versatility

If multiple specialized Machine Learning applications are developed for different vendors, then vendor-specific data requirements are met, but resource consumption and cost increase

Engineering Contradiction:
Improvevendor-specific data handlingVSAvoidcomputational resource consumption
Core Design Contradiction:
Adaptability or versatilityVSUse of energy by moving object

Solution Approach 1:

The patent creates a universal ML application infrastructure that handles multiple vendors through a common codebase and shared computational resources. The metadata-driven feature derivation engine enables a single application instance to adapt to different vendors dynamically, eliminating the need to deploy and maintain multiple separate applications, thereby reducing computational overhead and resource consumption.

Inventive Principle:
Principle #6Universality (Multi-functionality)

Solution Approach 2:

The patent utilizes parameter changes in the metadata schema to adapt the ML application to different vendors. By modifying metadata parameters such as column mappings, data type definitions, and feature derivation rules rather than changing the core application logic, the system can handle vendor-specific requirements efficiently without requiring separate application instances, thus optimizing resource utilization.

Inventive Principle:
Principle #35Parameter changes

3Reliability

If data cleaning and feature derivation are performed manually for each vendor, then data quality is improved, but the complexity and time required increase

Engineering Contradiction:
Improvedata qualityVSAvoidprocess complexity
Core Design Contradiction:
ReliabilityVSDevice complexity

Solution Approach 1:

The patent implements self-service automated data cleaning and feature derivation using metadata-driven processing. The system automatically discovers vendor-specific data characteristics through metadata inspection and performs appropriate cleaning and feature engineering operations without manual intervention. This automated approach maintains high data quality while significantly reducing process complexity compared to manual methods.

Inventive Principle:
Principle #25Self-service

Solution Approach 2:

The patent performs preliminary data cleaning and feature derivation automatically based on pre-defined metadata schemas and vendor-specific configuration files. By preparing and standardizing data in advance through automated preprocessing pipelines that leverage metadata information, the system ensures high data quality before ML model processing without requiring complex manual intervention at each vendor integration stage.

Inventive Principle:
Principle #10Preliminary action

Data Source

PatentEP3928467B1Managing telecommunication network event data
Publication Date: 2024.05.08 TELEFONAKTIEBOLAGET LM ERICSSON (PUBL)
  • EP3928467B1 patent drawingFigure 1
  • EP3928467B1 patent drawingFigure 2
  • EP3928467B1 patent drawingFigure 3

AI summary

A method (300) is disclosed for managing network event data in a telecommunication network, wherein the network event data is associated with products provided by different vendors. The method comprises obtaining a plurality of datasets comprising network event data associated with products provided by a different vendor, the network event data comprising a plurality of data instances representing a plurality of network event features (310). The method further comprises obtaining metadata describing the network event data in the datasets (320) and generating feature description vectors from the obtained metadata (330). The method further comprises calculating a metric of the relation between network event features represented in the datasets (340) and constructing a graph of the network event features represented in the datasets, with edges weighted according to the calculated metric (350). The method further comprises partitioning the graph into clusters by minimising an edge cut between network event features (360) and, for individual clusters in the graph, labelling network event features in the cluster with a normalised network event feature index (370). The method further comprises inputting the labelled network event data to a model for making at least one of recommendations or predictions (380).