Local Remote ML Device Classification Segmentation
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing telecommunications systems lack an efficient method for classifying devices operating in networks, which is crucial for monitoring quality of service (QoS) and managing network resources effectively.
Innovation Solution
A method and apparatus that utilize a machine learning classifier to classify devices operating in a network. The system queries a local device classification database and, if the device is unknown, uses basic information data to determine a local classifier device classification. This classification is then stored and updated using a remote ML device classifier based on historical network information.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If a local device classification database is used to store device classifications, then device classification accuracy is improved, but network storage resources and database maintenance complexity increase
Solution Approach 1:
The system segments device classification into two parts: a local device classification database for quick lookup and a remote machine learning classifier for comprehensive analysis. The local database stores frequently accessed classifications, while the remote system handles complex classification tasks, reducing local storage burden and maintenance complexity.
Solution Approach 2:
The system performs preliminary device classification using the local database before more complex remote classification is needed. This preliminary action filters out commonly classified devices, reducing the load on the remote system and minimizing database maintenance requirements while maintaining high classification accuracy.
2Adaptability or versatility
If a machine learning classifier is deployed to classify devices dynamically, then adaptability to changing device behaviors is improved, but computational resources and processing time increase
Solution Approach 1:
The system applies partial machine learning classification only when necessary - when devices are not found in the local database or when classification accuracy needs improvement. This partial action approach maintains adaptability to changing device behaviors while minimizing computational resource consumption by avoiding full ML processing for all devices.
Solution Approach 2:
The local device classification database acts as an intermediary between simple device identification and complex machine learning classification. It provides a fast, low-resource lookup mechanism that handles common cases, while the ML classifier serves as a backup for more complex scenarios, optimizing the balance between adaptability and resource usage.
3Measurement precision
If device classifications are updated frequently based on historical network information, then classification accuracy is improved, but network bandwidth consumption and system complexity increase
Solution Approach 1:
The system updates device classifications periodically based on historical network information rather than continuously. This periodic action allows the system to maintain accurate classifications by incorporating evolving device behaviors while controlling network bandwidth consumption by updating only at scheduled intervals rather than with every network event.
Data Source
AI summary
Aspects of the subject disclosure may include determining a local classifier device classification associated with a first device via a local machine learning (ML) device classifier according to basic information data associated with the first device, storing the local classifier device classification at a local device classification database, and storing the local classifier device classification at a device classification for ML training database, receiving a ML model update from a trainer for ML device classifier, the ML model update is generated by the trainer for ML device classifier according to the device classification for ML training database and a remote classifier device classification, and the remote classifier device classification is determined via a remote ML device classifier according to historical network information associated with the first device, and updating the local ML device classifier according to the ML model update. Other embodiments are disclosed.