Signal Quality Prediction Using Supervised Learning Models
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current methods for predicting signal and service quality for devices connected to radiofrequency antennas are based on statistical processing and cannot accurately forecast quality at a given time and position, failing to account for external factors like weather conditions and antenna saturation.
Innovation Solution
A method using supervised learning to train prediction models with data on signal and service quality parameters, device position, and time of connection, allowing for the estimation of quality decreases due to meteorological conditions and antenna saturation, enabling accurate predictions of signal and service quality.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Ease of manufacture
If statistical processing methods are used to establish coverage maps, then the method is simple and easy to implement, but the prediction accuracy of signal and service quality at a given time and position is insufficient
Solution Approach 1:
The patent transforms the approach from simple statistical processing to machine learning-based prediction by changing the parameters and methods used. It incorporates multiple input parameters including meteorological conditions, antenna saturation levels, device positions, and time of connection to generate accurate predictions of signal and service quality, thereby resolving the contradiction between implementation simplicity and prediction accuracy.
2Device complexity
If traditional coverage maps are used, then the system complexity is low, but the ability to predict quality at specific times and positions is insufficient
Solution Approach 1:
The patent implements preliminary training of machine learning models using historical data before actual prediction operations. The system pre-processes and stores training data including meteorological conditions, antenna saturation, and signal quality measurements, then trains prediction models in advance. This preliminary action enables reliable real-time predictions without requiring complex runtime computations, thus resolving the contradiction between system complexity and prediction reliability.
3Device complexity
If external factors like weather and antenna saturation are not considered, then the prediction model is simpler, but the accuracy of quality prediction is reduced
Solution Approach 1:
The patent segments the prediction model into distinct functional components: a first prediction model that predicts antenna saturation based on historical connection data, and a second prediction model that predicts signal and service quality based on the saturation prediction and meteorological conditions. This segmentation allows the system to handle multiple factors systematically without creating an overly complex monolithic model, thus resolving the contradiction between model complexity and prediction accuracy.
Data Source
AI summary
A method for predicting at least one parameter representative of a signal quality and/or service quality liable to be delivered to a device when it is connected to one radiofrequency antenna among a plurality of antennas that are configured to establish a connection with said device. The method includes: obtaining at least one parameter of decrease in the signal and/or service quality, predicting at least one parameter representative of a signal and/or service quality by applying at least one prediction model configured to predict at least one parameter representative of a signal quality and/or service quality, on the basis of an estimated parameter of decrease in the signal and/or service quality, and of an indication of a moment and of a position of the device for which the prediction must be carried out, the prediction model having been trained beforehand via supervised learning on a training database.