Cell Accessibility Prediction Using Ensemble KPI Models

Resolve Bottlenecks,
Find Innovative Solutions
Generate Solutions

Solution Overview

Problem

Existing mobile communication networks lack a mechanism to reliably measure, predict, and address cell accessibility issues, leading to user experience degradation and inefficient resource utilization due to delayed detection and manual resolution of connectivity problems.

Innovation Solution

A cell accessibility prediction system using ensemble machine learning models on KPIs like RACH, ERAB, and RRC success rates to proactively identify and automatically actuate corrective actions, reducing the time to resolve cell accessibility issues.

Engineering Contradictions & Design Principles

VSEngineering Contradiction Analysis

1Measurement precision

If machine learning models are trained on large datasets of cell images and features, then prediction accuracy improves, but computational resources and time requirements increase

Engineering Contradiction:
Improveprediction accuracyVSAvoidtraining time
Core Design Contradiction:
Measurement precisionVSLoss of time

Solution Approach 1:

The system performs preliminary actions by pre-training machine learning models on large datasets of cell images and features to establish baseline prediction capabilities. This pre-training phase captures general cell morphology patterns and staining characteristics, enabling the model to make reasonable predictions even with limited additional training data, thereby reducing the time required for subsequent specialized training while maintaining accuracy.

Inventive Principle:
Principle #10Preliminary action

Solution Approach 2:

The system applies local quality by focusing training efforts on specific, locally-relevant features rather than uniformly processing all cell data. The machine learning model identifies and prioritizes locally-important morphological features and staining patterns that are most predictive of cell accessibility, allowing accurate predictions with reduced computational resources by concentrating on discriminative local characteristics rather than processing entire cell images uniformly.

Inventive Principle:
Principle #3Local quality

2Reliability

If comprehensive cell morphology features are extracted and analyzed, then prediction reliability improves, but computational complexity increases

Engineering Contradiction:
Improveprediction reliabilityVSAvoidcomputational complexity
Core Design Contradiction:
ReliabilityVSDevice complexity

Solution Approach 1:

The system extracts only the most relevant and discriminative cell morphology features from comprehensive image data, rather than processing all available features. The machine learning model identifies and extracts key features such as nuclear shape, cytoplasmic texture, and staining intensity patterns that are most predictive of cell accessibility, discarding redundant information. This extraction approach maintains prediction reliability by focusing on critical features while reducing computational complexity by eliminating unnecessary processing of less informative data.

Inventive Principle:
Principle #2Taking out (Extraction)

3Reliability

If multiple machine learning models are trained on different feature sets, then prediction robustness improves, but system complexity increases

Engineering Contradiction:
Improveprediction robustnessVSAvoidsystem complexity
Core Design Contradiction:
ReliabilityVSDevice complexity

Solution Approach 1:

The system merges multiple machine learning models into a unified ensemble framework that processes different feature sets simultaneously. Rather than maintaining separate independent models, the system combines their predictions through weighted averaging or voting mechanisms, where each model contributes its specialized knowledge about different morphological features. This merging approach improves prediction robustness by leveraging diverse feature representations while reducing system complexity through a single integrated architecture that manages multiple feature processing pathways.

Inventive Principle:
Principle #5Merging (Combining)

Data Source

PatentEP4179762B1Cell accessibility prediction and actuation
Publication Date: 2026.04.29 TELEFONAKTIEBOLAGET LM ERICSSON (PUBL)
  • EP4179762B1 patent drawingFigure 1
  • EP4179762B1 patent drawingFigure 2
  • EP4179762B1 patent drawingFigure 3

AI summary

A method for predicting cell accessibility issues for a mobile network. The method includes receiving a set of metrics from the mobile network, processing a set of key performance indicators (KPIs) derived from the set of metrics in an ensemble machine learning model, the ensemble machine learning model including an RRC model, an RACH model, an ERAB model, and an S1 signaling model to generate at least one cell accessibility degradation prediction and a confidence score, and applying a root cause mapping to the at least one cell accessibility degradation prediction and the confidence score to identify at least one recommended action to correct a correlated cell accessibility issue.