Character-Embedded CNN for 5G/6G Protocol Log Classification
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing network testing technologies face inefficiencies in analyzing complex telecommunications software logs, requiring manual expert analysis that is error-prone and inefficient, while current AI solutions like LLMs struggle with large context windows and high computational costs.
Innovation Solution
A compact CNN architecture is developed for software log classification, utilizing a sequence-to-sequence model to initialize character embeddings and one-dimensional convolutional layers, capable of processing up to 200,000 characters, and is deployable on edge devices.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If LLM-based approaches are used for software log classification, then accuracy may be maintained, but computational cost and processing time increase significantly
Solution Approach 1:
The patent segments the log classification task into two stages: first using a lightweight CNN for rapid defect detection, and only then using more sophisticated methods if needed. This segmentation allows the system to achieve high accuracy for the majority of cases with minimal computational cost, reserving heavy processing only for ambiguous cases.
Solution Approach 2:
The patent employs a lightweight CNN model that acts as a 'cheap' first-line classifier that can be rapidly deployed and executed. This disposable-like approach uses a simple, computationally inexpensive model for the bulk of classification tasks, achieving 96% accuracy without the heavy computational burden of LLMs.
2Measurement precision
If manual expert analysis is used for software logs, then complex issues can be identified, but efficiency and productivity decrease
Solution Approach 1:
The patent implements a self-service automated classification system that processes software logs without requiring manual expert intervention. The CNN-based classifier autonomously analyzes logs, identifies defects, and classifies issues by protocol layer, achieving both high accuracy and scalability that manual analysis cannot match.
Solution Approach 2:
The patent replaces the mechanical process of manual expert analysis with an automated CNN-based system. This substitution maintains the ability to accurately identify complex issues while dramatically improving efficiency and productivity by eliminating human involvement in the classification process.
3Measurement precision
If comprehensive log analysis is performed to identify all defects, then measurement precision improves, but loss of time increases
Solution Approach 1:
The patent applies partial action by using a lightweight CNN that performs sufficient analysis to achieve 96% accuracy without attempting to analyze every possible aspect of the logs. This partial analysis approach achieves high precision while avoiding the time loss associated with exhaustive analysis methods.
Solution Approach 2:
The patent performs preliminary defect detection using the CNN classifier before more detailed analysis. This preliminary action quickly identifies obvious defects and classifies clear cases, reducing the time needed for comprehensive analysis by handling the majority of cases in advance.
Data Source
AI summary
In some implementations, a device may provide a software log to a convolutional neural network (CNN) associated with software log classification, wherein the CNN is associated with an embedding layer that is initialized with character embeddings extracted from a sequence-to-sequence model, and the CNN is associated with a block of one-dimensional convolutional layers that follows the character embeddings. The device may generate a software log classification using the CNN, wherein the software log classification indicates whether the software log is associated with an issue and a telecommunications protocol stack in which the issue or a defect has occurred.