Service Policy Optimization via Machine Learning

Resolve Bottlenecks,
Find Innovative SolutionsGenerate Solutions

Solution Overview

Problem

Ensuring customer satisfaction with services is challenging due to ever-changing conditions, such as sudden bandwidth demands from popular software applications, which existing service providers may not promptly address, leading to service congestion and customer complaints.

Innovation Solution

Deploying machine learning to monitor service data, detect anomalous patterns, and apply reinforcement learning to optimize resource allocation across multiple services, enabling quick responses to changing demands and improving overall service quality and customer satisfaction.

Engineering Contradictions & Design Principles

VSEngineering Contradiction Analysis

1Adaptability or versatility

If traditional service monitoring methods are used, then service providers can maintain basic service delivery, but they cannot promptly detect and respond to changing conditions such as sudden bandwidth demands

Engineering Contradiction:

Improveservice adaptation to changing conditionsVSAvoidresponse time to service issues

Core Design Contradiction:

Adaptability or versatilityVSLoss of time

Solution Approach 1:

The patent implements continuous feedback loops where service data is monitored, analyzed, and used to generate policies that are applied back to the service. Machine learning models process service metrics in real-time, detect anomalies, and trigger policy adjustments, creating a closed-loop system that rapidly adapts to changing conditions without manual intervention.

Inventive Principle:

Principle #23Feedback

Solution Approach 2:

The service management system operates autonomously by automatically monitoring service data, detecting anomalies through machine learning, generating optimization policies, and applying them without human intervention. The system self-adjusts resource allocation and service parameters in response to detected conditions, eliminating the need for manual service management and reducing response time.

Inventive Principle:

Principle #25Self-service

2Reliability

If machine learning models are trained separately for each service, then each service can be optimized independently, but the complexity of managing multiple models increases significantly

Engineering Contradiction:

Improveservice optimization effectivenessVSAvoidpolicy model management complexity

Core Design Contradiction:

ReliabilityVSDevice complexity

Solution Approach 1:

The patent employs a universal policy model framework that can process and optimize multiple different service types through a single unified system. The machine learning architecture is designed to handle diverse service data formats and optimization requirements, reducing the need for separate specialized models for each service while maintaining optimization effectiveness across all services.

Inventive Principle:

Principle #6Universality (Multi-functionality)

Solution Approach 2:

The patent combines multiple service optimization tasks into a unified policy management system. Instead of managing separate models for each service, the system merges the optimization functions into a single coordinated framework that shares resources, data processing capabilities, and policy management infrastructure, thereby reducing overall system complexity.

Inventive Principle:

Principle #5Merging (Combining)

3Productivity

If service policies are manually created and updated, then service providers maintain control over policy decisions, but the speed of policy implementation cannot keep pace with rapidly changing service conditions

Engineering Contradiction:

Improvepolicy implementation speedVSAvoidpolicy generation automation

Core Design Contradiction:

ProductivityVSExtent of automation

Solution Approach 1:

The patent replaces manual policy creation and updating processes with automated machine learning systems. Instead of human operators manually analyzing service data and creating policies, machine learning models automatically process service metrics, detect patterns, generate optimization policies, and implement them, substituting mechanical human processes with automated computational systems.

Inventive Principle:

Principle #28Mechanics substitution (Replace mechanical system)

Solution Approach 2:

The system performs preliminary analysis and policy generation in advance through continuous machine learning processing. By continuously monitoring service data and pre-computing optimization policies based on detected patterns, the system is prepared to rapidly implement policies when conditions change, eliminating the need for reactive manual policy creation.

Inventive Principle:

Principle #10Preliminary action

Data Source

PatentUS11328224B2Optimization of multiple services via machine learning

Publication Date: 2022.05.10 AT&T INTELLECTUAL PROPERTY I L P

AI summary

A method, computer-readable medium, and apparatus for modeling data of a service for providing a policy are disclosed. For example, a method may include a processor for generating a first policy for a first service by a first policy model using machine learning for processing first data of the first service, determining whether the first policy is to be applied to a second service, applying the first policy to the second service when the first policy is deemed to be applicable to the second service, wherein the applying the first policy provides the first policy to a second policy model using machine learning for processing second data of the second service, generating a second policy for the second service, and implementing the second policy in the second service, wherein the first service and the second service are provided by a single service provider.