Data Classification Threshold Adjustment for Rule Changes
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current data classification methods using machine learning technologies require extensive data, time, manpower, and computational resources, especially when classification rules change frequently, leading to inefficiencies and increased costs due to the need for retraining models and relabeling data.
Innovation Solution
An electronic apparatus and method that identify subtasks related to data classification, obtain scores for each data point based on a model, determine category information, and set threshold values using optimized threshold functions, allowing for flexible adaptation to changing rules without the need for extensive retraining or relabeling.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If machine learning models are retrained whenever classification rules change, then classification accuracy is maintained, but time consumption and computational cost increase significantly
Solution Approach 1:
The patent segments the classification task into a stable pre-trained model and a flexible threshold adjustment component. The model structure remains fixed while only the threshold parameters are updated when rules change, separating the heavy computational model from the lightweight rule adaptation layer.
Solution Approach 2:
Instead of retraining the entire model when rules change, the patent changes only the threshold parameters of the pre-trained model. This allows the model to adapt to new classification rules by adjusting threshold values rather than redistributing learned features.
2Reliability
If machine learning models are retrained whenever classification rules change, then classification accuracy is maintained, but computational resources and cost increase
Solution Approach 1:
The patent separates the computationally intensive model training phase from the lightweight threshold adjustment phase. Once the model is pre-trained on historical data, subsequent rule changes only require updating threshold parameters, not retraining the entire model architecture.
Solution Approach 2:
The patent uses parameter changes (threshold values) instead of model retraining to adapt to new classification rules. This approach maintains classification accuracy while dramatically reducing computational resource requirements compared to full model retraining.
3Reliability
If data is relabeled according to changed classification criteria, then model accuracy is maintained, but time and manpower requirements increase
Solution Approach 1:
The patent performs preliminary action by pre-training the model on historical data before rule changes occur. The model learns the underlying patterns once and then adapts to new rules by adjusting thresholds, eliminating the need for time-consuming relabeling of existing data.
Solution Approach 2:
Instead of relabeling data to reflect new classification rules, the patent changes the threshold parameters of the pre-trained model. This approach preserves the original labeled data while adapting the model's decision boundaries through parameter adjustment.
4Device complexity
If a single unified model is used for all classification subtasks, then device complexity is reduced, but adaptability to different classification rules decreases
Solution Approach 1:
The patent introduces dynamics by making the threshold parameters adjustable and adaptable. While the model architecture remains static and simple, the threshold parameters can be dynamically modified to adapt to different classification rules and policies without changing the underlying model structure.
Data Source
AI summary
Provided is a method of setting criteria for classification of data in an electronic apparatus, the method including identifying a subtask related to data classification, obtaining a plurality of data, based on a model associated with the subtask, obtaining a score associated with the subtask for each of the plurality of data, identifying category information associated with the subtask for each of the plurality of data, and based on the category information and the score, determining a threshold value corresponding to the subtask.