Predictive Feedback Loop for Online Application Configuration

The patent transitions from static SLA thresholds to dynamic QoE prediction models that continuously adapt to changing network conditions and application characteristics. The system uses machine learning models that are trained on historical data and updated in real-time to accurately predict user experience metrics such as MOS (Mean Opinion Score) for voice quality, allowing the network management system to respond dynamically rather than relying on fixed thresholds.

The patent implements feedback mechanisms where actual user experience measurements are fed back into the machine learning models to refine predictions. The system collects data from network elements and application users, compares it with predicted values, and uses this feedback to retrain and improve the accuracy of QoE prediction models, creating a continuous improvement loop that enhances prediction precision over time.

The patent introduces machine learning prediction models as intermediary components that translate complex network metrics into meaningful user experience predictions. Instead of directly measuring difficult-to-obtain user experience data, the system uses network metrics (bandwidth, latency, packet loss) as inputs to the prediction models, which then output QoE estimates. This intermediary approach maintains measurement simplicity while improving reliability by capturing the complex relationship between network conditions and actual user experience.

The patent changes the parameters used for SLA compliance assessment from traditional network metrics to predicted user experience metrics. The system transforms physical network parameters into psychological/perceptual parameters through machine learning models, allowing SLA compliance to be judged based on actual user experience quality rather than just network performance indicators. This parameter transformation resolves the contradiction by maintaining ease of network metric collection while achieving more reliable user experience-based SLA compliance.

The patent applies preliminary action by having the machine learning model predict optimal application parameters before network conditions deteriorate or before the application needs to adapt. The system proactively determines the best configuration settings (such as video bitrate, audio codec selection, or packet size) based on predicted network conditions and provides these recommendations to the application in advance, allowing the application to pre-configure itself for optimal performance without requiring complex real-time decision-making logic.

The patent enables application self-service by providing the application with automated parameter recommendations from the machine learning model. Instead of requiring complex manual configuration or sophisticated internal adaptation logic, the application can simply receive suggested parameter values from the network system and apply them automatically. This maintains operational simplicity for the application while achieving high adaptability to varying network conditions through the intelligence of the prediction model.