KQI Prediction via Association Rule Learning
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Measuring and improving customer quality of service (QoS) in telecommunications networks is challenging due to the difficulty and cost of obtaining key quality indicators (KQIs), while key performance indicators (KPIs) are less expensive but time-consuming to calculate, making it hard to identify QoS performance levels effectively.
Innovation Solution
The proposed solution employs association rule learning to generate data rules by categorizing KQIs and KPIs into groups and calculating association frequencies, allowing for the prediction of KQIs from KPIs and diagnosis of network anomalies, even when direct KQI measurements are costly or impractical.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If direct KQI measurements are used to accurately determine quality of service levels, then measurement precision is improved, but cost and time consumption increase significantly
Solution Approach 1:
The patent creates a predictive model that copies the relationship between KPIs and KQIs by training on historical data. Once trained, the model generates predicted KQI values from current KPI measurements, serving as a lightweight copy of the expensive direct KQI measurement process. This allows operators to obtain QoS estimates quickly without performing time-consuming direct KQI calculations.
Solution Approach 2:
The patent introduces KPIs as intermediary variables that mediate between easily measurable network parameters and the desired QoS outcomes. By establishing predictive relationships through association rule learning, KPIs serve as intermediaries that translate cheap, fast measurements into accurate QoS predictions, bridging the gap between operational data and quality assessments.
2Measurement precision
If direct KQI measurements are performed to accurately identify service quality levels, then measurement precision is improved, but cost increases
Solution Approach 1:
The patent replaces expensive direct KQI measurements with cheap KPI-based predictions. The system uses readily available KPI data that costs minimal resources to collect and processes through the trained predictive model, generating accurate QoS estimates at a fraction of the cost of direct measurement. This disposable approach uses inexpensive surrogate measurements instead of costly primary measurements.
Solution Approach 2:
The trained predictive model serves as a reusable copy of the expensive measurement process. Once the model is trained on historical KQI data, it can be deployed indefinitely to generate predictions from cheap KPI inputs, eliminating the need to repeatedly incur the high costs of direct KQI measurements while maintaining measurement accuracy.
3Measurement precision
If comprehensive KQI data collection is performed to improve service quality analysis, then measurement precision is improved, but device complexity increases
Solution Approach 1:
The patent extracts only the essential predictive relationships from the complex web of KQI and KPI data. Through association rule learning, the system identifies and extracts the most significant correlations between specific KPIs and KQIs, building a streamlined predictive model that captures essential dynamics without requiring processing of all possible data combinations. This reduces computational complexity while preserving predictive accuracy.
Solution Approach 2:
The patent segments the complex data analysis task into distinct phases: (1) data collection and preprocessing, (2) association rule learning and model training, and (3) real-time prediction. By dividing the overall process into manageable segments, the system handles complexity in controlled stages rather than attempting to process all data relationships simultaneously, making the system more tractable and efficient.
Data Source
AI summary
The disclosure relates to technology for processing data sets to generate data rules for the data sets in a communications network. A first set of data including key quality indicators (KQIs) indicative of a quality of service and a second set of data including key performance indicators (KPIs) indicative of a performance level are received. The first data set and the second data set are categorized using a first value into a plurality of KQI groups and a second value into a plurality of KPI groups, respectively. Each of the KQI and KPI groups are identified with a label. Each of the KQI and KPI groups identified with a same label are processed by application of association rule learning to generate the data rules. The data rules model a relationship between the KQIs and the KPIs by calculating association frequencies.